0 00:00:00,000 --> 00:00:30,000 Dear viewer, these subtitles were generated by a machine via the service Trint and therefore are (very) buggy. If you are capable, please help us to create good quality subtitles: https://c3subtitles.de/talk/674 Thanks! 1 00:00:15,370 --> 00:00:17,799 So today, 2 00:00:17,800 --> 00:00:20,409 it's my very pleasure 3 00:00:20,410 --> 00:00:23,439 to introduce Matt Bernard 4 00:00:23,440 --> 00:00:24,820 and Alex Halderman. 5 00:00:26,030 --> 00:00:27,049 Met is a Ph.D. 6 00:00:27,050 --> 00:00:29,029 student, and Alex is a professor at the 7 00:00:29,030 --> 00:00:30,739 University of Michigan. 8 00:00:30,740 --> 00:00:32,749 Both of them do security research, and 9 00:00:32,750 --> 00:00:35,359 they have a focus on e-voting and 10 00:00:35,360 --> 00:00:37,909 its security implications. 11 00:00:37,910 --> 00:00:39,229 Today, they're going to talk about 12 00:00:39,230 --> 00:00:41,479 Recount 2016 and uninvited 13 00:00:41,480 --> 00:00:43,849 security audits of the US 14 00:00:43,850 --> 00:00:45,349 presidential election. 15 00:00:45,350 --> 00:00:47,779 Please give them a very warm welcome and 16 00:00:47,780 --> 00:00:48,949 a huge round of applause. 17 00:00:48,950 --> 00:00:49,950 Thank you. 18 00:01:04,440 --> 00:01:05,440 Hmm. 19 00:01:13,170 --> 00:01:14,170 There we go. 20 00:01:20,150 --> 00:01:21,169 Hello, everyone. 21 00:01:21,170 --> 00:01:22,399 I'm Alex Halderman. 22 00:01:22,400 --> 00:01:24,559 It's wonderful to be back here 23 00:01:24,560 --> 00:01:26,719 at the CCC. 24 00:01:26,720 --> 00:01:28,879 I've spoken in recent years 25 00:01:28,880 --> 00:01:31,039 about other areas of my work, 26 00:01:31,040 --> 00:01:33,469 like cryptographic attacks 27 00:01:33,470 --> 00:01:35,629 on TLS, ZMapp, an 28 00:01:35,630 --> 00:01:37,639 internet wide scanning. 29 00:01:37,640 --> 00:01:39,829 But today we're 30 00:01:39,830 --> 00:01:41,689 coming to an issue that has been for 31 00:01:41,690 --> 00:01:43,819 many, many years, very central 32 00:01:43,820 --> 00:01:46,069 to my heart, and that is ensuring 33 00:01:46,070 --> 00:01:48,799 the integrity of elections 34 00:01:48,800 --> 00:01:51,019 in 2016 in the United 35 00:01:51,020 --> 00:01:52,009 States. 36 00:01:52,010 --> 00:01:54,229 We've seen some unprecedented things 37 00:01:54,230 --> 00:01:56,539 during the presidential election, 38 00:01:56,540 --> 00:01:59,239 and I and Matt and others 39 00:01:59,240 --> 00:02:01,489 have had the great privilege of taking 40 00:02:01,490 --> 00:02:03,829 part in what is essentially 41 00:02:03,830 --> 00:02:05,899 our own kind of hack of 42 00:02:05,900 --> 00:02:07,549 this election. 43 00:02:07,550 --> 00:02:09,859 A last ditch attempt to find 44 00:02:09,860 --> 00:02:11,959 out for sure whether 45 00:02:11,960 --> 00:02:15,079 the election outcome was genuine. 46 00:02:15,080 --> 00:02:17,689 This is our first time speaking 47 00:02:17,690 --> 00:02:19,909 in front of giving a full length 48 00:02:19,910 --> 00:02:22,069 talk about the recount 49 00:02:22,070 --> 00:02:24,019 process and what we found. 50 00:02:24,020 --> 00:02:26,059 And I can think of no better place to 51 00:02:26,060 --> 00:02:28,250 give it than here at the CCC. 52 00:02:35,110 --> 00:02:37,359 Thanks. And thank you, Alex. 53 00:02:37,360 --> 00:02:39,519 So as some of you may have heard, we 54 00:02:39,520 --> 00:02:40,899 recently had a presidential election in 55 00:02:40,900 --> 00:02:41,900 the United States. 56 00:02:43,420 --> 00:02:45,549 It was a contest largely between Hillary 57 00:02:45,550 --> 00:02:46,809 Clinton and Donald Trump. 58 00:02:46,810 --> 00:02:48,519 And as you probably know, Donald Trump 59 00:02:48,520 --> 00:02:49,689 won. 60 00:02:49,690 --> 00:02:51,429 What you may not have known is that this 61 00:02:51,430 --> 00:02:52,479 was kind of a surprise. 62 00:02:53,500 --> 00:02:54,999 FiveThirtyEight, which is a US election 63 00:02:55,000 --> 00:02:57,099 blog, had Hillary Clinton 64 00:02:57,100 --> 00:02:58,659 winning seventy one point four percent of 65 00:02:58,660 --> 00:02:59,949 the time. 66 00:02:59,950 --> 00:03:01,929 The New York Times gave her an 85 percent 67 00:03:01,930 --> 00:03:03,039 chance. 68 00:03:03,040 --> 00:03:05,139 Sam Wang at Princeton gave her a greater 69 00:03:05,140 --> 00:03:07,269 than 99 percent chance and promise 70 00:03:07,270 --> 00:03:08,680 to eat a bug if she lost. 71 00:03:10,450 --> 00:03:12,519 And just generally, the general consensus 72 00:03:12,520 --> 00:03:14,649 was that Hillary Clinton was 73 00:03:14,650 --> 00:03:15,939 going to win this election. 74 00:03:17,230 --> 00:03:18,189 So what happened? 75 00:03:18,190 --> 00:03:20,769 As many of you probably know, 76 00:03:20,770 --> 00:03:21,789 she didn't win the election. 77 00:03:22,870 --> 00:03:25,089 And here we have Sam 78 00:03:25,090 --> 00:03:27,039 Wang on national television getting a 79 00:03:27,040 --> 00:03:28,040 bug. 80 00:03:31,360 --> 00:03:33,490 So what did the result of this look like? 81 00:03:34,540 --> 00:03:36,189 Trump won the Electoral College in the 82 00:03:36,190 --> 00:03:37,599 United States. We aren't a direct 83 00:03:37,600 --> 00:03:39,399 democracy. We have this thing called the 84 00:03:39,400 --> 00:03:42,099 Electoral College, where every state gets 85 00:03:42,100 --> 00:03:43,839 a proportional amount of votes based on 86 00:03:43,840 --> 00:03:46,059 its population and the election 87 00:03:46,060 --> 00:03:47,889 is decided in most states, it's the 88 00:03:47,890 --> 00:03:48,999 majority. 89 00:03:49,000 --> 00:03:50,259 The candidate that wins the majority of 90 00:03:50,260 --> 00:03:51,609 votes in that state gets all of their 91 00:03:51,610 --> 00:03:54,039 electoral votes, with two exceptions. 92 00:03:54,040 --> 00:03:56,739 So Trump won 304 electoral votes. 93 00:03:56,740 --> 00:03:58,599 Clinton won 227. 94 00:03:58,600 --> 00:04:00,819 However, Clinton won 48 percent 95 00:04:00,820 --> 00:04:03,099 of the popular vote and beat Trump 96 00:04:03,100 --> 00:04:04,929 by three million votes. 97 00:04:04,930 --> 00:04:06,249 So this gives you an idea of how the 98 00:04:06,250 --> 00:04:07,419 Electoral College can sort of, 99 00:04:08,470 --> 00:04:10,240 and it causes some interaction. 100 00:04:11,350 --> 00:04:13,749 Donald Trump, of course, insists that he 101 00:04:13,750 --> 00:04:16,059 won the Electoral College in a landslide, 102 00:04:17,140 --> 00:04:19,088 which isn't quite true. 103 00:04:19,089 --> 00:04:21,729 In fact, he had the 48th 104 00:04:21,730 --> 00:04:23,619 best victory in the Electoral College in 105 00:04:23,620 --> 00:04:25,749 U.S. history, right behind Harry S. 106 00:04:25,750 --> 00:04:26,889 Truman. 107 00:04:26,890 --> 00:04:29,379 So a little overblown 108 00:04:29,380 --> 00:04:30,309 there. 109 00:04:30,310 --> 00:04:31,859 But what happened? 110 00:04:31,860 --> 00:04:33,939 There are sort of two explanations here 111 00:04:33,940 --> 00:04:35,649 why? Why were we so wrong? 112 00:04:35,650 --> 00:04:37,449 Why was this such a surprise? 113 00:04:37,450 --> 00:04:38,739 One is that the polls are just 114 00:04:38,740 --> 00:04:40,179 systematically wrong. Nate Silver at 115 00:04:40,180 --> 00:04:42,309 FiveThirtyEight, Sam Wang and 116 00:04:42,310 --> 00:04:43,539 many others have just said, You know, 117 00:04:43,540 --> 00:04:45,219 there there is high uncertainty, higher 118 00:04:45,220 --> 00:04:46,479 uncertainty in the polls than we 119 00:04:46,480 --> 00:04:47,859 anticipated. 120 00:04:47,860 --> 00:04:50,409 And this seems reasonable. 121 00:04:50,410 --> 00:04:52,059 The other explanation is that the 122 00:04:52,060 --> 00:04:54,399 elections was rigged, which is 123 00:04:54,400 --> 00:04:56,649 a position that prior to the election was 124 00:04:56,650 --> 00:04:59,049 was largely supported by President 125 00:04:59,050 --> 00:05:00,050 elect Trump. 126 00:05:00,850 --> 00:05:03,129 This is probably unlikely, but 127 00:05:03,130 --> 00:05:05,529 nevertheless was rigging feasible. 128 00:05:05,530 --> 00:05:07,659 How would we even know, you 129 00:05:07,660 --> 00:05:09,249 know, how are we going to determine 130 00:05:09,250 --> 00:05:10,479 whether it was just the polls were wrong 131 00:05:10,480 --> 00:05:11,739 or whether or not there was actually 132 00:05:11,740 --> 00:05:13,389 rigging? So with that, I'll give it back 133 00:05:13,390 --> 00:05:14,390 to Alex here. 134 00:05:22,670 --> 00:05:25,369 So elections in the United States 135 00:05:25,370 --> 00:05:27,769 are really quite complicated 136 00:05:27,770 --> 00:05:29,629 to pull off logistically. 137 00:05:29,630 --> 00:05:32,269 We have a massive scale with about 200 138 00:05:32,270 --> 00:05:34,129 million registered voters. 139 00:05:34,130 --> 00:05:36,499 Elections are highly distributed because 140 00:05:36,500 --> 00:05:38,629 we operate in a federal system. 141 00:05:38,630 --> 00:05:40,549 Most of the work of running an election 142 00:05:40,550 --> 00:05:42,919 happens at the state, the county, even 143 00:05:42,920 --> 00:05:45,139 the local municipal level, 144 00:05:45,140 --> 00:05:47,629 and there's a lot of complexity to 145 00:05:47,630 --> 00:05:48,829 the process. 146 00:05:48,830 --> 00:05:51,649 Our ballots are sometimes ridiculously 147 00:05:51,650 --> 00:05:54,019 complicated, loaded with local races 148 00:05:54,020 --> 00:05:55,490 unique to each jurisdiction. 149 00:05:56,570 --> 00:05:58,789 And above all, our 150 00:05:58,790 --> 00:06:01,669 elections are very sensitive to latency. 151 00:06:01,670 --> 00:06:03,949 So whatever we use to count them, 152 00:06:03,950 --> 00:06:05,959 we need to get the results basically on 153 00:06:05,960 --> 00:06:07,849 election night, where people just find it 154 00:06:07,850 --> 00:06:09,319 unacceptable. 155 00:06:09,320 --> 00:06:10,879 So in the in the country, there were more 156 00:06:10,880 --> 00:06:13,369 than 13000 different voting 157 00:06:13,370 --> 00:06:15,559 jurisdictions that are making decisions 158 00:06:15,560 --> 00:06:17,959 about how to implement elections. 159 00:06:17,960 --> 00:06:20,209 They're running elections on election 160 00:06:20,210 --> 00:06:23,269 night in more than 100 87000 161 00:06:23,270 --> 00:06:26,179 voting locations or precincts. 162 00:06:26,180 --> 00:06:28,039 And our technology? 163 00:06:28,040 --> 00:06:30,499 Well, we have 52 different models 164 00:06:30,500 --> 00:06:32,929 of voting machines in the US, 165 00:06:32,930 --> 00:06:35,539 and the choice of which to use 166 00:06:35,540 --> 00:06:38,089 is left up to the states, the counties. 167 00:06:38,090 --> 00:06:40,369 Even in some cases, individual cities. 168 00:06:41,660 --> 00:06:42,709 All right, so here is a map. 169 00:06:42,710 --> 00:06:44,389 Just to illustrate a little bit of this 170 00:06:44,390 --> 00:06:46,579 diversity, you see all kinds of different 171 00:06:46,580 --> 00:06:47,479 systems. 172 00:06:47,480 --> 00:06:50,149 Optical scan diary paper ballots 173 00:06:50,150 --> 00:06:52,639 used in different parts of the country. 174 00:06:52,640 --> 00:06:54,169 Here's just one state. 175 00:06:54,170 --> 00:06:56,449 This is Arkansas, and you can see just 176 00:06:56,450 --> 00:06:59,269 within a single state, the variety can be 177 00:06:59,270 --> 00:07:00,470 extremely large. 178 00:07:01,700 --> 00:07:03,439 Here's one of those long, complicated 179 00:07:03,440 --> 00:07:04,729 ballots I mentioned. 180 00:07:04,730 --> 00:07:07,789 This comes from Dallas, Texas. 181 00:07:07,790 --> 00:07:09,799 This year, it's a an example of the 182 00:07:09,800 --> 00:07:12,289 ballot. This is just page one 183 00:07:12,290 --> 00:07:14,059 asking you for your party and your 184 00:07:14,060 --> 00:07:15,889 presidential choice. 185 00:07:15,890 --> 00:07:17,089 Here's the second page. 186 00:07:18,580 --> 00:07:20,259 And here are the six other pages, 187 00:07:23,140 --> 00:07:25,719 so if you can imagine trying to count 188 00:07:25,720 --> 00:07:28,449 a ballot like this by hand 189 00:07:28,450 --> 00:07:30,999 would be extremely time consuming 190 00:07:31,000 --> 00:07:32,409 because you'd have to go through each 191 00:07:32,410 --> 00:07:33,669 individual race. 192 00:07:33,670 --> 00:07:36,399 The fastest way to count ballots 193 00:07:36,400 --> 00:07:38,799 is just to sort them by who was selected 194 00:07:38,800 --> 00:07:41,079 on a particular race and 195 00:07:41,080 --> 00:07:42,579 count the stacks. 196 00:07:42,580 --> 00:07:44,589 Instead, you'd have to go through each of 197 00:07:44,590 --> 00:07:46,749 the 175 or however, many 198 00:07:46,750 --> 00:07:49,269 questions on this ballot individually, 199 00:07:49,270 --> 00:07:50,739 and you can see some of them or even a 200 00:07:50,740 --> 00:07:52,839 paragraph long describing a 201 00:07:52,840 --> 00:07:54,310 particular referendum. 202 00:07:56,970 --> 00:07:59,159 So we use two primary styles 203 00:07:59,160 --> 00:08:01,259 of voting machines that those 52 204 00:08:01,260 --> 00:08:03,389 models of election technology mostly 205 00:08:03,390 --> 00:08:05,999 fall within these two styles. 206 00:08:06,000 --> 00:08:08,249 These are optical scan voting, where 207 00:08:08,250 --> 00:08:10,349 the voter fills out a paper ballot 208 00:08:10,350 --> 00:08:12,719 and runs it through a machine that scans 209 00:08:12,720 --> 00:08:14,339 it right in front of them and spits it 210 00:08:14,340 --> 00:08:17,339 back out if the voter has made a mistake. 211 00:08:17,340 --> 00:08:19,169 Remember how complicated these ballots 212 00:08:19,170 --> 00:08:21,749 are? We need the computers to help us. 213 00:08:21,750 --> 00:08:24,089 The second style is called direct 214 00:08:24,090 --> 00:08:26,239 recording electronic or D.R 215 00:08:26,240 --> 00:08:27,179 e-voting. 216 00:08:27,180 --> 00:08:29,279 And in this style of machine, 217 00:08:29,280 --> 00:08:31,769 the voter casts a vote on a computer 218 00:08:31,770 --> 00:08:33,839 terminal. The computer records 219 00:08:33,840 --> 00:08:36,239 the vote in an electronic memory, 220 00:08:36,240 --> 00:08:38,609 and sometimes, but not always, 221 00:08:38,610 --> 00:08:40,769 the computer prints a piece of paper 222 00:08:40,770 --> 00:08:42,928 that the voter can see that is saved 223 00:08:42,929 --> 00:08:45,179 for later. It makes a physical record 224 00:08:45,180 --> 00:08:46,529 of their vote. 225 00:08:46,530 --> 00:08:49,079 And for years, many scientists, 226 00:08:49,080 --> 00:08:51,749 including me and including Rob, 227 00:08:51,750 --> 00:08:53,699 have been criticizing DORRIE voting 228 00:08:53,700 --> 00:08:55,799 machines in particular because they're 229 00:08:55,800 --> 00:08:57,119 fully electronic. 230 00:08:57,120 --> 00:08:59,099 Anyone who tampered with their software 231 00:08:59,100 --> 00:09:01,110 can change every record of the vote. 232 00:09:02,220 --> 00:09:03,689 Let me give you an example of that. 233 00:09:03,690 --> 00:09:05,969 This is one of the Dorrie voting machine 234 00:09:05,970 --> 00:09:07,289 models used in the U.S. 235 00:09:07,290 --> 00:09:10,139 a machine called the Diebold AQ votes 236 00:09:10,140 --> 00:09:12,419 that 10 years ago was the most widely 237 00:09:12,420 --> 00:09:15,059 used voting machine in the country. 238 00:09:15,060 --> 00:09:17,699 I was part of a study that obtained 239 00:09:17,700 --> 00:09:20,039 one of these machines from an anonymous 240 00:09:20,040 --> 00:09:22,109 source in reverse engineered 241 00:09:22,110 --> 00:09:24,209 it to do an uninvited security 242 00:09:24,210 --> 00:09:25,589 audit. 243 00:09:25,590 --> 00:09:27,929 What we found was really 244 00:09:27,930 --> 00:09:29,639 pretty devastating. 245 00:09:29,640 --> 00:09:30,749 One. 246 00:09:30,750 --> 00:09:33,119 These machines well, they use a removable 247 00:09:33,120 --> 00:09:35,669 memory card before the election. 248 00:09:35,670 --> 00:09:37,769 Most voting machines do, and this 249 00:09:37,770 --> 00:09:39,989 is to load the design of the ballot 250 00:09:39,990 --> 00:09:41,729 into the machine. 251 00:09:41,730 --> 00:09:44,039 After the election, the card is removed. 252 00:09:44,040 --> 00:09:45,869 It has a copy of the election results, 253 00:09:45,870 --> 00:09:47,849 and that's what's counted on election 254 00:09:47,850 --> 00:09:49,049 night. 255 00:09:49,050 --> 00:09:50,789 Well, the first thing we found was that 256 00:09:50,790 --> 00:09:53,189 anyone who had temporary access 257 00:09:53,190 --> 00:09:55,289 to that removable memory card could 258 00:09:55,290 --> 00:09:57,299 corrupt the files on it in a way that 259 00:09:57,300 --> 00:09:59,489 would exploit buffer overflows 260 00:09:59,490 --> 00:10:01,349 and other vulnerabilities in the voting 261 00:10:01,350 --> 00:10:03,449 machine so that when the card 262 00:10:03,450 --> 00:10:06,239 was inserted before the election, malware 263 00:10:06,240 --> 00:10:08,279 would be installed onto the machine. 264 00:10:10,020 --> 00:10:12,419 The second thing we found 265 00:10:12,420 --> 00:10:14,099 was that that malware. 266 00:10:14,100 --> 00:10:16,349 Well, let me show you, 267 00:10:16,350 --> 00:10:18,329 here's an election we ran as a 268 00:10:18,330 --> 00:10:20,759 demonstration on our machine. 269 00:10:20,760 --> 00:10:22,979 George Washington versus Benedict Arnold, 270 00:10:22,980 --> 00:10:24,960 the traitor of the American Revolution. 271 00:10:25,980 --> 00:10:28,169 Well, on our machine infected with our 272 00:10:28,170 --> 00:10:30,479 malware. No matter how many people vote 273 00:10:30,480 --> 00:10:31,919 for George Washington. 274 00:10:31,920 --> 00:10:34,259 Benedict Arnold always wins. 275 00:10:34,260 --> 00:10:35,909 That's because our malware is just 276 00:10:35,910 --> 00:10:37,589 another application running on the 277 00:10:37,590 --> 00:10:39,839 computer, and it can access all 278 00:10:39,840 --> 00:10:41,429 the same records of the vote that the 279 00:10:41,430 --> 00:10:43,559 software running the election can. 280 00:10:43,560 --> 00:10:45,689 So every time someone votes, 281 00:10:45,690 --> 00:10:47,669 our malware goes and modifies those 282 00:10:47,670 --> 00:10:49,529 records, the electronic records of the 283 00:10:49,530 --> 00:10:51,839 vote to make sure our preferred candidate 284 00:10:51,840 --> 00:10:53,999 wins. It even has a nice UI. 285 00:10:54,000 --> 00:10:56,129 Here you can pick who wants, who wins and 286 00:10:56,130 --> 00:10:57,130 by how much. 287 00:10:59,250 --> 00:11:01,529 Anyway, real malware would likely be more 288 00:11:01,530 --> 00:11:02,530 surreptitious. 289 00:11:04,580 --> 00:11:05,659 So are U.S. 290 00:11:05,660 --> 00:11:07,639 voting machines secure? 291 00:11:07,640 --> 00:11:09,799 Well, we know in this one case 292 00:11:09,800 --> 00:11:11,929 there's one theory that the machine 293 00:11:11,930 --> 00:11:13,189 was not. 294 00:11:13,190 --> 00:11:15,289 But what about other machines? 295 00:11:15,290 --> 00:11:17,389 Well, many different models have 296 00:11:17,390 --> 00:11:19,489 been studied by academics 297 00:11:19,490 --> 00:11:21,229 and other independent security 298 00:11:21,230 --> 00:11:22,489 researchers. 299 00:11:22,490 --> 00:11:24,739 And in every case, 300 00:11:24,740 --> 00:11:26,839 the machines have been found to be 301 00:11:26,840 --> 00:11:29,029 susceptible to malware infection from 302 00:11:29,030 --> 00:11:31,219 their memory cards that can lead 303 00:11:31,220 --> 00:11:33,289 to the machine giving 304 00:11:33,290 --> 00:11:35,689 incorrect results like 305 00:11:35,690 --> 00:11:37,609 this one. A few years ago, I turned into 306 00:11:37,610 --> 00:11:39,709 a Pac-Man machine because there wasn't 307 00:11:39,710 --> 00:11:41,059 any more science to do. 308 00:11:43,250 --> 00:11:45,919 But it's not just the door is 309 00:11:45,920 --> 00:11:46,920 one way. 310 00:11:48,890 --> 00:11:51,019 It's not just these diary voting 311 00:11:51,020 --> 00:11:53,509 machines that are susceptible to malware. 312 00:11:53,510 --> 00:11:56,029 The optical scan machines can be infected 313 00:11:56,030 --> 00:11:58,219 too, and their electronic vote 314 00:11:58,220 --> 00:12:00,499 records can also be tampered with in just 315 00:12:00,500 --> 00:12:02,629 the same way so that the results 316 00:12:02,630 --> 00:12:04,399 that are reported electronically on 317 00:12:04,400 --> 00:12:06,529 election night are wrong. 318 00:12:06,530 --> 00:12:08,779 The paper will still be right, but 319 00:12:08,780 --> 00:12:10,999 the electronic results, the ones we hear 320 00:12:11,000 --> 00:12:13,009 about immediately will be wrong. 321 00:12:13,010 --> 00:12:14,629 And this has been true for just about 322 00:12:14,630 --> 00:12:16,759 every model of optical scan machine 323 00:12:16,760 --> 00:12:18,199 as well. 324 00:12:18,200 --> 00:12:19,549 In fact, every U.S. 325 00:12:19,550 --> 00:12:21,619 voting machine that's been subjected 326 00:12:21,620 --> 00:12:24,439 to rigorous independent security testing 327 00:12:24,440 --> 00:12:26,539 has been found to be susceptible 328 00:12:26,540 --> 00:12:28,609 to the use of malware spreading 329 00:12:28,610 --> 00:12:30,799 on the memory cards to introduce 330 00:12:30,800 --> 00:12:32,269 a vote stealing attack. 331 00:12:35,370 --> 00:12:36,370 So 332 00:12:37,590 --> 00:12:39,569 how would we use this to hack an 333 00:12:39,570 --> 00:12:40,570 election? 334 00:12:43,830 --> 00:12:45,059 So there are three main ways that we 335 00:12:45,060 --> 00:12:46,469 could hack an election. 336 00:12:46,470 --> 00:12:48,659 We could alter election night results. 337 00:12:48,660 --> 00:12:50,979 This would undermine the credibility of 338 00:12:50,980 --> 00:12:53,249 of the election, even if especially 339 00:12:53,250 --> 00:12:55,319 if it was detected or obviously it 340 00:12:55,320 --> 00:12:56,879 could get the wrong person elected. 341 00:12:56,880 --> 00:12:58,199 We could perform denial of service 342 00:12:58,200 --> 00:12:59,339 attack. We could prevent people from 343 00:12:59,340 --> 00:13:00,689 voting. We can make the lines at the 344 00:13:00,690 --> 00:13:01,889 polls really long. 345 00:13:01,890 --> 00:13:03,539 We could prevent votes from being 346 00:13:03,540 --> 00:13:04,559 counted. 347 00:13:04,560 --> 00:13:05,759 And then there's general political 348 00:13:05,760 --> 00:13:07,769 interference. Let's say that, you know, a 349 00:13:07,770 --> 00:13:09,239 candidate that I don't like is running. 350 00:13:09,240 --> 00:13:11,549 I can go about a cyber 351 00:13:11,550 --> 00:13:13,859 campaign to basically ruin 352 00:13:13,860 --> 00:13:15,479 their their chances. 353 00:13:15,480 --> 00:13:16,649 And we've seen all three of these in 354 00:13:16,650 --> 00:13:19,649 practice in 2014 in Ukraine. 355 00:13:19,650 --> 00:13:22,349 There was a massive cyber attack, one 356 00:13:22,350 --> 00:13:24,569 one part of which was to try and 357 00:13:24,570 --> 00:13:26,249 change the vote totals. 358 00:13:26,250 --> 00:13:27,569 And another part of which was to prevent 359 00:13:27,570 --> 00:13:29,399 the votes from being counted at all. 360 00:13:29,400 --> 00:13:31,679 And in this election, we've seen 361 00:13:31,680 --> 00:13:33,989 multiple cyber attacks against political 362 00:13:33,990 --> 00:13:35,819 campaigns, specifically the Democratic 363 00:13:35,820 --> 00:13:38,219 National Convention, voter registration 364 00:13:38,220 --> 00:13:40,349 databases in Illinois, Arizona 365 00:13:40,350 --> 00:13:41,999 and a few other states. 366 00:13:42,000 --> 00:13:43,499 And then John Podesta, the campaign 367 00:13:43,500 --> 00:13:45,239 manager for Hillary Clinton, also had his 368 00:13:45,240 --> 00:13:46,240 emails hacked. 369 00:13:47,040 --> 00:13:49,259 And in almost all of these cases, 370 00:13:49,260 --> 00:13:51,299 the US government has formally attributed 371 00:13:51,300 --> 00:13:52,829 these attacks to Russia. 372 00:13:52,830 --> 00:13:54,449 So we know that these kinds of attacks 373 00:13:54,450 --> 00:13:55,450 happen. 374 00:13:56,190 --> 00:13:57,419 But what about other attacks? 375 00:13:57,420 --> 00:13:58,459 What about invisible attacks? 376 00:13:58,460 --> 00:13:59,519 So let's say we want to influence the 377 00:13:59,520 --> 00:14:00,520 election in some other way 378 00:14:02,460 --> 00:14:03,899 to do this. There are really three main 379 00:14:03,900 --> 00:14:06,029 challenges we need to 380 00:14:06,030 --> 00:14:08,069 overcome the diverse, decentralized 381 00:14:08,070 --> 00:14:10,289 technology that Alex just mentioned. 382 00:14:10,290 --> 00:14:11,819 But this is pretty easy because of the 383 00:14:11,820 --> 00:14:13,319 Electoral College. 384 00:14:13,320 --> 00:14:15,269 Elections very frequently come down to 385 00:14:15,270 --> 00:14:16,320 very few states 386 00:14:17,340 --> 00:14:19,439 and a very small margin of votes. 387 00:14:19,440 --> 00:14:20,999 To give you an idea in this election, you 388 00:14:21,000 --> 00:14:22,889 would need to flip the result in two 389 00:14:22,890 --> 00:14:24,689 different states, both of which had 390 00:14:24,690 --> 00:14:26,909 margins within one percent. 391 00:14:26,910 --> 00:14:27,989 So that's not really that big of a 392 00:14:27,990 --> 00:14:28,990 challenge. 393 00:14:29,790 --> 00:14:31,409 The second challenge is that how do you 394 00:14:31,410 --> 00:14:32,459 infect the machines? 395 00:14:32,460 --> 00:14:34,049 And when you ask poll workers this in an 396 00:14:34,050 --> 00:14:35,999 election officials their first responses, 397 00:14:36,000 --> 00:14:37,499 the machines are not connected to the 398 00:14:37,500 --> 00:14:40,109 internet, as I'm sure many of you know, 399 00:14:40,110 --> 00:14:42,269 this doesn't really matter. 400 00:14:42,270 --> 00:14:44,579 All voting machines are programed through 401 00:14:44,580 --> 00:14:45,719 centralized election management 402 00:14:45,720 --> 00:14:47,609 computers, oftentimes which are connected 403 00:14:47,610 --> 00:14:48,610 to the internet. 404 00:14:49,320 --> 00:14:51,479 And so all you have to do 405 00:14:51,480 --> 00:14:52,889 is put a virus on this machine and then 406 00:14:52,890 --> 00:14:54,659 every memory card that is programed with 407 00:14:54,660 --> 00:14:56,489 that computer, every voting machine it 408 00:14:56,490 --> 00:14:59,039 goes into will become infected. 409 00:14:59,040 --> 00:15:01,739 So you effectively have a county level 410 00:15:01,740 --> 00:15:03,899 hack that it's, you know, pretty easy 411 00:15:03,900 --> 00:15:05,039 to do so. 412 00:15:06,500 --> 00:15:09,209 But how likely is this 413 00:15:09,210 --> 00:15:11,339 in the United States? Very many, 414 00:15:11,340 --> 00:15:13,469 many jurisdictions outsource ballot 415 00:15:13,470 --> 00:15:15,629 programing to these small businesses. 416 00:15:15,630 --> 00:15:17,489 For instance, in Michigan, 75 percent of 417 00:15:17,490 --> 00:15:19,559 counties use two 418 00:15:19,560 --> 00:15:21,239 companies, both of which have fewer than 419 00:15:21,240 --> 00:15:22,229 20 employees. 420 00:15:22,230 --> 00:15:23,759 This is governmental business solutions, 421 00:15:23,760 --> 00:15:25,049 one of them. 422 00:15:25,050 --> 00:15:26,249 And as you can see, you know, they don't 423 00:15:26,250 --> 00:15:28,080 even have be enabled on their website. 424 00:15:29,850 --> 00:15:31,139 They're pretty small. 425 00:15:31,140 --> 00:15:32,639 Like I said, a small business, jobs 426 00:15:32,640 --> 00:15:33,959 services, I believe four or five 427 00:15:33,960 --> 00:15:35,699 different states, and this is their 428 00:15:35,700 --> 00:15:38,099 warehouse. So you can see some active TSX 429 00:15:38,100 --> 00:15:39,149 is there. 430 00:15:39,150 --> 00:15:41,399 And, you know, basically they're pretty 431 00:15:41,400 --> 00:15:42,989 public facing. It's pretty small. 432 00:15:44,340 --> 00:15:46,259 It wouldn't be that hard to to infect 433 00:15:46,260 --> 00:15:48,029 machines at a very high rate. 434 00:15:49,380 --> 00:15:51,479 The third challenge is these paper 435 00:15:51,480 --> 00:15:52,499 records. 436 00:15:52,500 --> 00:15:53,489 Seventy percent of U.S. 437 00:15:53,490 --> 00:15:54,599 votes have paper records, 438 00:15:56,160 --> 00:15:58,440 and this paper is sort of akin to 439 00:15:59,490 --> 00:16:01,829 in most critical infrastructure systems. 440 00:16:01,830 --> 00:16:04,169 We want to have some sort of failsafe 441 00:16:04,170 --> 00:16:06,329 in case the electronics go down. 442 00:16:06,330 --> 00:16:07,709 So, for instance, in airplanes, they have 443 00:16:07,710 --> 00:16:09,359 physical compasses in case for some 444 00:16:09,360 --> 00:16:10,619 reason the navigation system doesn't 445 00:16:10,620 --> 00:16:11,699 work. 446 00:16:11,700 --> 00:16:13,199 Paper records are effectively the same 447 00:16:13,200 --> 00:16:14,099 thing. 448 00:16:14,100 --> 00:16:16,169 The major caveats to paper are they 449 00:16:16,170 --> 00:16:17,819 are slow and inexpensive, Detalhes Alex 450 00:16:17,820 --> 00:16:19,109 mentioned, whereas memory is fast and 451 00:16:19,110 --> 00:16:20,110 cheap. 452 00:16:20,460 --> 00:16:22,199 But the major strength is that every 453 00:16:22,200 --> 00:16:24,089 paper record is verified by the voter. 454 00:16:24,090 --> 00:16:25,529 Either it's an optical scan ballot that 455 00:16:25,530 --> 00:16:26,849 the voter fills out and checks before 456 00:16:26,850 --> 00:16:28,109 they feed it into the machine. 457 00:16:28,110 --> 00:16:30,029 Or it's a paper audit trail produced by a 458 00:16:30,030 --> 00:16:31,709 diary that the voter can look at before 459 00:16:31,710 --> 00:16:33,119 they push the big red button that says, 460 00:16:33,120 --> 00:16:35,759 Cast your vote, whereas on memory cards, 461 00:16:35,760 --> 00:16:37,019 there's no way to verify it at all. 462 00:16:37,020 --> 00:16:38,549 It's just, you know, regular computer 463 00:16:38,550 --> 00:16:39,550 memory. 464 00:16:40,440 --> 00:16:42,389 So really, what we need to do is we need 465 00:16:42,390 --> 00:16:44,999 to audit the memory cards with 466 00:16:45,000 --> 00:16:45,919 paper records. 467 00:16:45,920 --> 00:16:47,939 And this would trivially defeat most 468 00:16:47,940 --> 00:16:48,909 attacks. 469 00:16:48,910 --> 00:16:51,089 Unfortunately, as I mentioned, 470 00:16:51,090 --> 00:16:52,169 70 percent of states 471 00:16:53,220 --> 00:16:56,039 are recorded votes on paper in 2016. 472 00:16:56,040 --> 00:16:57,929 This is up significantly and is largely 473 00:16:57,930 --> 00:17:00,029 due to efforts by Alex and several 474 00:17:00,030 --> 00:17:02,069 other people over the last 10 or so 475 00:17:02,070 --> 00:17:03,389 years. 476 00:17:03,390 --> 00:17:04,919 Talking about election integrity and ways 477 00:17:04,920 --> 00:17:06,118 to prevent fraud. 478 00:17:07,920 --> 00:17:08,920 But unfortunately, 479 00:17:10,050 --> 00:17:12,179 unfortunately, most states never look 480 00:17:12,180 --> 00:17:14,779 at the paper and 481 00:17:14,780 --> 00:17:16,919 you know you have this great, you know, 482 00:17:16,920 --> 00:17:18,539 way to defend against an attack, but you 483 00:17:18,540 --> 00:17:19,540 never use it. 484 00:17:20,430 --> 00:17:22,769 So in short, hacking 485 00:17:22,770 --> 00:17:24,689 an election is actually much, much easier 486 00:17:24,690 --> 00:17:25,690 than you would think. 487 00:17:29,430 --> 00:17:30,419 All right. 488 00:17:30,420 --> 00:17:32,429 So let's put this together, how would you 489 00:17:32,430 --> 00:17:34,349 actually hack a U.S. 490 00:17:34,350 --> 00:17:36,869 presidential election? 491 00:17:36,870 --> 00:17:39,659 Well, step one before the election. 492 00:17:39,660 --> 00:17:42,029 Use pre-election polls like Nate Silver 493 00:17:42,030 --> 00:17:44,129 provides to identify the states 494 00:17:44,130 --> 00:17:45,929 that are likely to be closest that are 495 00:17:45,930 --> 00:17:48,839 going to be within a percent or so. 496 00:17:48,840 --> 00:17:50,909 Step two target some large 497 00:17:50,910 --> 00:17:53,189 counties where their service providers 498 00:17:53,190 --> 00:17:55,139 and compromise their election management 499 00:17:55,140 --> 00:17:56,699 system computers. 500 00:17:56,700 --> 00:17:58,229 I'll leave it as an exercise to the 501 00:17:58,230 --> 00:18:00,959 attacker to find out how to compromise 502 00:18:00,960 --> 00:18:03,239 the election management system by, say, 503 00:18:03,240 --> 00:18:04,730 starting by emailing Sue. 504 00:18:07,830 --> 00:18:10,409 Step three Use the compromised 505 00:18:10,410 --> 00:18:12,599 election management system to spread your 506 00:18:12,600 --> 00:18:14,399 infection to the individual voting 507 00:18:14,400 --> 00:18:15,419 machines. 508 00:18:15,420 --> 00:18:17,189 Developing an attack for one of these 509 00:18:17,190 --> 00:18:19,859 machines is not terribly difficult. 510 00:18:19,860 --> 00:18:22,019 I and others have done it again and again 511 00:18:22,020 --> 00:18:23,219 in the laboratory. 512 00:18:23,220 --> 00:18:24,659 All you need to do is buy one. 513 00:18:24,660 --> 00:18:26,729 Government surplus on eBay to test 514 00:18:26,730 --> 00:18:27,730 it out. 515 00:18:28,380 --> 00:18:30,509 Finally, step four Your 516 00:18:30,510 --> 00:18:33,119 attack steals votes on the computer, 517 00:18:33,120 --> 00:18:34,889 but no one ever looks at the paper 518 00:18:34,890 --> 00:18:36,179 because this will show you. 519 00:18:36,180 --> 00:18:38,489 Most states just throw away that paper 520 00:18:38,490 --> 00:18:39,839 without ever looking at it. 521 00:18:41,780 --> 00:18:44,659 So coming back to 2016 522 00:18:44,660 --> 00:18:46,879 and the presidential election, 523 00:18:46,880 --> 00:18:49,369 we knew on November, 524 00:18:49,370 --> 00:18:51,649 on November 8th, excuse me, not December 525 00:18:51,650 --> 00:18:54,259 8th, on November 8th on Election Day, 526 00:18:54,260 --> 00:18:56,539 that hacking was possible. 527 00:18:56,540 --> 00:18:57,979 And at the end of the day, when the 528 00:18:57,980 --> 00:19:00,109 election results came in, we 529 00:19:00,110 --> 00:19:02,839 saw that the result was extremely close 530 00:19:02,840 --> 00:19:05,299 was surprising compared 531 00:19:05,300 --> 00:19:06,589 to the polling. 532 00:19:06,590 --> 00:19:08,689 We knew at that point that there had 533 00:19:08,690 --> 00:19:10,849 been cyber attacks of an unprecedented 534 00:19:10,850 --> 00:19:13,249 nature in American politics aimed 535 00:19:13,250 --> 00:19:14,929 previously at interfering with the 536 00:19:14,930 --> 00:19:17,449 election. This was all before Election 537 00:19:17,450 --> 00:19:20,299 Day, and we knew that it was feasible 538 00:19:20,300 --> 00:19:22,609 for an attacker to change votes 539 00:19:22,610 --> 00:19:24,949 on enough machines to have stolen 540 00:19:24,950 --> 00:19:25,970 the election result. 541 00:19:27,290 --> 00:19:29,659 Shockingly, at least shockingly to me 542 00:19:29,660 --> 00:19:31,939 and many other people, even under 543 00:19:31,940 --> 00:19:34,849 these circumstances, approximately 544 00:19:34,850 --> 00:19:36,049 zero U.S. 545 00:19:36,050 --> 00:19:38,089 states, we're going to look at enough 546 00:19:38,090 --> 00:19:40,009 paper ballots to know whether the 547 00:19:40,010 --> 00:19:41,359 computers had been hacked. 548 00:19:42,520 --> 00:19:44,859 This is a major gap 549 00:19:44,860 --> 00:19:46,569 in our system. 550 00:19:46,570 --> 00:19:48,729 I previously had believed that the paper 551 00:19:48,730 --> 00:19:50,979 would provide a fairly strong deterrent, 552 00:19:50,980 --> 00:19:53,079 but if even in 2016, 553 00:19:53,080 --> 00:19:54,549 we're not going to look at any of the 554 00:19:54,550 --> 00:19:55,749 paper? 555 00:19:55,750 --> 00:19:57,160 Well, it might as well not be there. 556 00:19:58,860 --> 00:20:01,259 So at this point, there were five 557 00:20:01,260 --> 00:20:03,449 weeks between Election Day on November 558 00:20:03,450 --> 00:20:05,669 8th and December 13th, 559 00:20:05,670 --> 00:20:07,829 which is a deadline under federal law for 560 00:20:07,830 --> 00:20:10,199 states to lock in their Electoral College 561 00:20:10,200 --> 00:20:11,339 votes. 562 00:20:11,340 --> 00:20:13,799 If there was no, if there was 563 00:20:13,800 --> 00:20:15,899 fraud that had taken place, it had 564 00:20:15,900 --> 00:20:18,059 to be exposed by this date in order to 565 00:20:18,060 --> 00:20:20,489 have any hope of changing the outcome 566 00:20:20,490 --> 00:20:21,490 of the election. 567 00:20:22,840 --> 00:20:25,149 So what to do under these circumstances, 568 00:20:25,150 --> 00:20:27,219 I and other election integrity 569 00:20:27,220 --> 00:20:29,349 advocates got together and 570 00:20:29,350 --> 00:20:31,359 started discussing some possibilities, 571 00:20:31,360 --> 00:20:32,769 but we didn't really have a good 572 00:20:32,770 --> 00:20:35,799 solution. Was there any way possible 573 00:20:35,800 --> 00:20:38,049 to make the state actually examine 574 00:20:38,050 --> 00:20:39,909 the physical evidence they had in a way 575 00:20:39,910 --> 00:20:41,829 that could potentially detect a cyber 576 00:20:41,830 --> 00:20:42,830 attack? 577 00:20:44,400 --> 00:20:46,559 So the first question and really 578 00:20:46,560 --> 00:20:48,329 the beginning of this whole process for 579 00:20:48,330 --> 00:20:50,489 me was on the 13th 580 00:20:50,490 --> 00:20:52,589 on the Sunday after Election Day, I was 581 00:20:52,590 --> 00:20:55,019 on a flight from New York to Los Angeles 582 00:20:55,020 --> 00:20:57,119 when I got a frantic email from 583 00:20:57,120 --> 00:20:58,769 Barbara Simons, another election 584 00:20:58,770 --> 00:21:00,419 integrity advocate. 585 00:21:00,420 --> 00:21:02,639 The email said, Can you join 586 00:21:02,640 --> 00:21:04,769 a phone call in 15 minutes? 587 00:21:04,770 --> 00:21:06,839 It's with the Clinton campaign. 588 00:21:06,840 --> 00:21:08,880 Hillary herself might be on the call. 589 00:21:10,180 --> 00:21:12,249 Well, what do 590 00:21:12,250 --> 00:21:14,439 I do, I'm on a flight, 591 00:21:14,440 --> 00:21:16,689 so I find a way to work 592 00:21:16,690 --> 00:21:18,939 around the airplane, which flies 593 00:21:18,940 --> 00:21:21,009 attempts to block Skype and so forth 594 00:21:21,010 --> 00:21:23,019 and find a way to dial into the call. 595 00:21:29,940 --> 00:21:31,529 It doesn't work well enough for me to 596 00:21:31,530 --> 00:21:33,659 talk on this first call, but I can listen 597 00:21:33,660 --> 00:21:35,789 in and I can hear what they're saying. 598 00:21:35,790 --> 00:21:37,859 Clinton herself isn't there, but 599 00:21:37,860 --> 00:21:39,929 Podesta, her campaign manager and 600 00:21:39,930 --> 00:21:42,479 many other top people from the campaign 601 00:21:42,480 --> 00:21:44,849 are on and they're listening very, very 602 00:21:44,850 --> 00:21:45,839 carefully. 603 00:21:45,840 --> 00:21:48,089 As Barbara and others are describing 604 00:21:48,090 --> 00:21:50,339 arguments why Clinton should 605 00:21:50,340 --> 00:21:52,469 recount should request a recount 606 00:21:52,470 --> 00:21:54,180 in several key states. 607 00:21:55,290 --> 00:21:57,359 So candidates under American law have 608 00:21:57,360 --> 00:22:00,149 this ability. They can demand a recount 609 00:22:00,150 --> 00:22:02,309 depending on the state, as 610 00:22:02,310 --> 00:22:04,829 long as they pay the requisite costs. 611 00:22:04,830 --> 00:22:07,109 But wouldn't Clinton do it? 612 00:22:07,110 --> 00:22:08,969 So there were a few concerns that the 613 00:22:08,970 --> 00:22:11,129 campaign brought up the first one, and 614 00:22:11,130 --> 00:22:13,199 the most aggravating to me is 615 00:22:13,200 --> 00:22:15,119 they wanted to know what evidence we had 616 00:22:15,120 --> 00:22:17,639 that the election had been hacked. 617 00:22:17,640 --> 00:22:18,869 Well, wait a minute. That seems 618 00:22:18,870 --> 00:22:20,579 completely backwards. 619 00:22:20,580 --> 00:22:22,499 The purpose of a recount of any 620 00:22:22,500 --> 00:22:24,599 examination of the paper ballots is 621 00:22:24,600 --> 00:22:26,549 to gather evidence that the outcome was 622 00:22:26,550 --> 00:22:29,159 correct or not without 623 00:22:29,160 --> 00:22:30,419 looking at them. 624 00:22:30,420 --> 00:22:32,309 We don't have the evidence to begin with. 625 00:22:32,310 --> 00:22:34,019 All we can do is say that it would be 626 00:22:34,020 --> 00:22:36,329 possible for an attack to have changed 627 00:22:36,330 --> 00:22:38,249 the result without leaving any visible 628 00:22:38,250 --> 00:22:39,479 evidence. 629 00:22:39,480 --> 00:22:41,339 Another concern was how to pay for the 630 00:22:41,340 --> 00:22:43,469 recounts. The cost was likely to rise to 631 00:22:43,470 --> 00:22:45,539 the millions of dollars, but may be 632 00:22:45,540 --> 00:22:48,209 most important and most aggravating 633 00:22:48,210 --> 00:22:49,319 was they were concerned about 634 00:22:49,320 --> 00:22:51,509 backpedaling from the concession, 635 00:22:51,510 --> 00:22:53,639 about tip toeing away from having 636 00:22:53,640 --> 00:22:55,769 conceded the victory to Donald 637 00:22:55,770 --> 00:22:57,899 Trump and for Clinton in 638 00:22:57,900 --> 00:22:58,919 2016. 639 00:22:58,920 --> 00:23:00,959 This was a particular concern because 640 00:23:00,960 --> 00:23:02,879 there was this famous incident during the 641 00:23:02,880 --> 00:23:05,159 third presidential debate, 642 00:23:05,160 --> 00:23:07,229 where Trump refused to 643 00:23:07,230 --> 00:23:09,389 say yes, he would accept the election 644 00:23:09,390 --> 00:23:10,390 results. 645 00:23:11,370 --> 00:23:13,439 And Clinton reacted to this by 646 00:23:13,440 --> 00:23:15,509 saying that something truly horrifying. 647 00:23:15,510 --> 00:23:17,999 It's a direct threat to our democracy. 648 00:23:18,000 --> 00:23:19,349 Well, there are two things he could have 649 00:23:19,350 --> 00:23:21,629 meant, right? One was he wouldn't accept 650 00:23:21,630 --> 00:23:23,979 the result when settled right, 651 00:23:23,980 --> 00:23:25,859 that he wouldn't concede power to the 652 00:23:25,860 --> 00:23:28,229 victor. That is truly horrifying. 653 00:23:28,230 --> 00:23:29,579 But the other, which is kind of a 654 00:23:29,580 --> 00:23:31,619 rational thing, is, well, I don't want to 655 00:23:31,620 --> 00:23:33,989 accept the results until all the evidence 656 00:23:33,990 --> 00:23:35,879 is examined to make sure it wasn't 657 00:23:35,880 --> 00:23:37,289 rigged. 658 00:23:37,290 --> 00:23:39,419 Well, she was committed to a position 659 00:23:39,420 --> 00:23:41,879 that not accepting the election 660 00:23:41,880 --> 00:23:43,949 result immediately was a 661 00:23:43,950 --> 00:23:45,449 threat to democracy. 662 00:23:45,450 --> 00:23:47,609 And so any kind of tiptoeing away from 663 00:23:47,610 --> 00:23:49,799 that would have been a political problem. 664 00:23:51,150 --> 00:23:52,589 All right. So that was the first 665 00:23:52,590 --> 00:23:54,119 conversation. We had a number of 666 00:23:54,120 --> 00:23:56,069 conversations trying to convince the 667 00:23:56,070 --> 00:23:57,989 Clinton people to move ahead with 668 00:23:57,990 --> 00:24:00,059 recounts, but it didn't 669 00:24:00,060 --> 00:24:02,009 seem like we were ever going to get 670 00:24:02,010 --> 00:24:03,010 there. 671 00:24:03,390 --> 00:24:05,519 But then we had a new idea, 672 00:24:05,520 --> 00:24:07,439 this time from David Jefferson, another 673 00:24:07,440 --> 00:24:09,719 election integrity advocate. 674 00:24:09,720 --> 00:24:12,209 His idea was is wonderful. 675 00:24:12,210 --> 00:24:15,059 Insight was that actually in many states, 676 00:24:15,060 --> 00:24:17,549 any candidate can request a recount. 677 00:24:17,550 --> 00:24:19,199 And although you might think we're a two 678 00:24:19,200 --> 00:24:21,449 party system in the U.S., there 679 00:24:21,450 --> 00:24:22,469 are third parties. 680 00:24:23,730 --> 00:24:26,219 In fact, we identified one third party. 681 00:24:26,220 --> 00:24:29,069 Jill Stein, the Green Party candidate. 682 00:24:29,070 --> 00:24:31,259 And Jill, we had a conversation 683 00:24:31,260 --> 00:24:32,939 with her, culminating in a talk on the 684 00:24:32,940 --> 00:24:35,699 twenty second where she 685 00:24:35,700 --> 00:24:38,729 unbelievably was willing to help. 686 00:24:38,730 --> 00:24:41,429 Jill has a long history of 687 00:24:41,430 --> 00:24:43,769 being a champion of election integrity. 688 00:24:43,770 --> 00:24:46,049 She was on the ballot in most states. 689 00:24:46,050 --> 00:24:48,029 She ended up winning about one percent of 690 00:24:48,030 --> 00:24:50,249 the vote. Not really all that much, 691 00:24:50,250 --> 00:24:52,109 but still enough. 692 00:24:52,110 --> 00:24:53,969 And most importantly, she was willing to 693 00:24:53,970 --> 00:24:54,929 help. 694 00:24:54,930 --> 00:24:57,059 This was an absolutely pivotal moment 695 00:24:57,060 --> 00:24:58,200 for the entire effort. 696 00:24:59,460 --> 00:25:01,529 But then that same day, at 697 00:25:01,530 --> 00:25:03,809 about six o'clock in the evening, we also 698 00:25:03,810 --> 00:25:06,029 had what may have been the biggest 699 00:25:06,030 --> 00:25:07,349 setback. 700 00:25:07,350 --> 00:25:09,599 And that is in New York magazine. 701 00:25:09,600 --> 00:25:11,279 There was a story published about the 702 00:25:11,280 --> 00:25:13,049 conversations that we were having with 703 00:25:13,050 --> 00:25:15,299 the Clinton campaign, and this 704 00:25:15,300 --> 00:25:17,429 was based on information that had 705 00:25:17,430 --> 00:25:19,469 been provided by to the press in 706 00:25:19,470 --> 00:25:21,659 violation of everyone's confidence by 707 00:25:21,660 --> 00:25:23,189 a person who was there. 708 00:25:23,190 --> 00:25:25,049 And it was a second hand account that got 709 00:25:25,050 --> 00:25:27,119 most of the facts wrong. 710 00:25:27,120 --> 00:25:29,429 In fact, it quoted it said that I 711 00:25:29,430 --> 00:25:31,829 and others had found persuasive evidence 712 00:25:31,830 --> 00:25:34,439 that the results may have been hacked. 713 00:25:34,440 --> 00:25:36,479 Well, wait a minute. That's 180 degrees 714 00:25:36,480 --> 00:25:38,579 the opposite. We want to check and see 715 00:25:38,580 --> 00:25:40,289 if there's evidence that the results have 716 00:25:40,290 --> 00:25:42,419 been hacked. We know it's a possibility, 717 00:25:42,420 --> 00:25:43,950 but that's all at this point. 718 00:25:45,120 --> 00:25:47,459 What to do if this is the 719 00:25:47,460 --> 00:25:49,559 thing that I have to respond to in 720 00:25:49,560 --> 00:25:51,389 my phone started ringing off the hook at 721 00:25:51,390 --> 00:25:53,429 that moment and continued to ring off the 722 00:25:53,430 --> 00:25:55,079 hook for weeks. 723 00:25:55,080 --> 00:25:57,239 In fact, I just checked my voicemail 724 00:25:57,240 --> 00:25:59,309 yesterday and I have a call a month ago 725 00:25:59,310 --> 00:26:01,409 from my congresswoman. 726 00:26:01,410 --> 00:26:03,450 I'll call you back on Monday, Debbie. 727 00:26:05,880 --> 00:26:06,960 So what to do? 728 00:26:08,010 --> 00:26:10,979 So what I did to try to 729 00:26:10,980 --> 00:26:13,109 not have to answer the press in 730 00:26:13,110 --> 00:26:14,909 a way that says, Well, we don't really 731 00:26:14,910 --> 00:26:15,869 have any evidence. 732 00:26:15,870 --> 00:26:18,029 End of story was I stayed up all 733 00:26:18,030 --> 00:26:20,219 night with Matt and others and wrote 734 00:26:20,220 --> 00:26:22,079 this medium piece that some of you may 735 00:26:22,080 --> 00:26:24,389 have read in which I present 736 00:26:24,390 --> 00:26:26,489 the case for recounts 737 00:26:26,490 --> 00:26:28,479 that the only way to know whether an. 738 00:26:28,480 --> 00:26:30,579 Hack occurred is to look at the 739 00:26:30,580 --> 00:26:32,289 evidence, and if we're not going to look 740 00:26:32,290 --> 00:26:34,389 at the evidence in 2016 by 741 00:26:34,390 --> 00:26:36,729 looking at the paper, when will we ever? 742 00:26:36,730 --> 00:26:38,430 A candidate needs to act now. 743 00:26:41,250 --> 00:26:43,589 So that kicked off what really 744 00:26:43,590 --> 00:26:46,049 has been an ongoing media circus 745 00:26:46,050 --> 00:26:47,549 around the recounts. 746 00:26:47,550 --> 00:26:49,829 And unfortunately, this narrative of 747 00:26:49,830 --> 00:26:52,049 is there some kind of secret evidence 748 00:26:52,050 --> 00:26:53,050 or not? 749 00:26:54,000 --> 00:26:55,829 Has been very difficult to get around, 750 00:26:55,830 --> 00:26:57,539 but I think we're doing our best. 751 00:26:57,540 --> 00:26:59,939 And in this talk, we'll show you why 752 00:26:59,940 --> 00:27:02,159 the recounts were essential and 753 00:27:02,160 --> 00:27:03,989 what evidence the recounts are able to 754 00:27:03,990 --> 00:27:06,119 provide vis-a-vis the election 755 00:27:06,120 --> 00:27:07,120 result. 756 00:27:07,800 --> 00:27:09,869 So what happened next 757 00:27:09,870 --> 00:27:12,629 on the 23rd, the day before Thanksgiving, 758 00:27:12,630 --> 00:27:15,239 the same day my medium post went live. 759 00:27:15,240 --> 00:27:17,309 Jill Stein announced that she was 760 00:27:17,310 --> 00:27:19,619 going to lead an effort to recount three 761 00:27:19,620 --> 00:27:21,839 critical states in 762 00:27:21,840 --> 00:27:23,429 just two days. 763 00:27:23,430 --> 00:27:25,379 A crowdfunding campaign that Jill and the 764 00:27:25,380 --> 00:27:27,659 Green Party led raised more 765 00:27:27,660 --> 00:27:29,969 than $5 million for the recounts. 766 00:27:29,970 --> 00:27:31,889 That's more money than she raised for her 767 00:27:31,890 --> 00:27:33,780 entire presidential campaign. 768 00:27:41,400 --> 00:27:43,489 160000 individual 769 00:27:43,490 --> 00:27:45,829 donors. It's absolutely incredible 770 00:27:45,830 --> 00:27:47,899 how many people care passionately 771 00:27:47,900 --> 00:27:49,819 about the integrity of our elections. 772 00:27:51,230 --> 00:27:53,569 The next question now we have the money, 773 00:27:53,570 --> 00:27:56,119 we have a candidate where to recount. 774 00:28:01,380 --> 00:28:02,639 So in order to figure out where to 775 00:28:02,640 --> 00:28:04,379 recount, we decided to sort of take 776 00:28:04,380 --> 00:28:06,059 attacked like I mentioned earlier, if I 777 00:28:06,060 --> 00:28:08,159 were an attacker, where would I want to, 778 00:28:08,160 --> 00:28:09,629 you know, where would I hack to have the 779 00:28:09,630 --> 00:28:10,799 most result for the least amount of 780 00:28:10,800 --> 00:28:11,800 effort? 781 00:28:12,240 --> 00:28:14,309 And as I said, there are a few states in 782 00:28:14,310 --> 00:28:16,229 the United States that had very small 783 00:28:16,230 --> 00:28:18,449 margins. Three of these states were 784 00:28:18,450 --> 00:28:20,460 Wisconsin, Michigan and Pennsylvania. 785 00:28:21,660 --> 00:28:23,219 All three of them are projected to be won 786 00:28:23,220 --> 00:28:24,989 by Clinton. All three of them are won by 787 00:28:24,990 --> 00:28:27,299 Trump. They combined for forty six 788 00:28:27,300 --> 00:28:29,459 electoral votes. The electoral margin was 789 00:28:29,460 --> 00:28:31,799 37 votes. I believe 790 00:28:31,800 --> 00:28:33,989 all three had margins, less than eight 791 00:28:33,990 --> 00:28:35,519 tenths of a percent. 792 00:28:35,520 --> 00:28:37,739 And in total, the US election was 793 00:28:37,740 --> 00:28:39,509 possibly decided by the seventy eight 794 00:28:39,510 --> 00:28:41,639 thousand votes in these three states. 795 00:28:41,640 --> 00:28:43,259 The seventy eight thousand margin. 796 00:28:44,610 --> 00:28:46,409 And moreover, we had to find states where 797 00:28:46,410 --> 00:28:48,179 we can actually do something. 798 00:28:48,180 --> 00:28:49,709 Michigan and Wisconsin allow candidates 799 00:28:49,710 --> 00:28:51,389 to petition for recounts. 800 00:28:51,390 --> 00:28:52,919 Pennsylvania requires a court order. 801 00:28:52,920 --> 00:28:54,149 But we thought our chances were pretty 802 00:28:54,150 --> 00:28:55,150 good. 803 00:28:55,680 --> 00:28:57,779 So in Wisconsin, to start 804 00:28:57,780 --> 00:28:59,849 us off, the margin was 805 00:28:59,850 --> 00:29:01,349 just under twenty three thousand votes. 806 00:29:01,350 --> 00:29:03,239 That's about point seven point eight 807 00:29:03,240 --> 00:29:04,439 percent. 808 00:29:04,440 --> 00:29:06,209 The technology there is almost entirely 809 00:29:06,210 --> 00:29:08,549 off scan, and every ballot that is cast 810 00:29:08,550 --> 00:29:10,799 in Wisconsin has a paper record. 811 00:29:10,800 --> 00:29:12,479 And as I mentioned the law, any candidate 812 00:29:12,480 --> 00:29:14,339 can demand a recount if he or she can pay 813 00:29:14,340 --> 00:29:16,499 the cost, and each county 814 00:29:16,500 --> 00:29:18,029 in Wisconsin decides how to do the 815 00:29:18,030 --> 00:29:18,929 recount. You can. 816 00:29:18,930 --> 00:29:20,429 You can recount all the votes by hand 817 00:29:20,430 --> 00:29:22,049 counting all the paper, or you could just 818 00:29:22,050 --> 00:29:24,209 run the ballots back through a machine. 819 00:29:24,210 --> 00:29:25,889 And so in Wisconsin, that decision is 820 00:29:25,890 --> 00:29:28,529 left up to the counties in Michigan. 821 00:29:28,530 --> 00:29:30,089 The margin was much smaller was about 822 00:29:30,090 --> 00:29:33,449 11000 votes point two or three percent. 823 00:29:33,450 --> 00:29:35,309 And by the way, Michigan is also where we 824 00:29:35,310 --> 00:29:36,690 live. So it was sort of convenient. 825 00:29:38,220 --> 00:29:39,899 The technology there is entirely optical 826 00:29:39,900 --> 00:29:41,219 scan. 827 00:29:41,220 --> 00:29:43,649 And again, all votes use paper, and 828 00:29:43,650 --> 00:29:45,209 the law states that any aggrieved 829 00:29:45,210 --> 00:29:47,639 candidate can file can pay a statutory 830 00:29:47,640 --> 00:29:49,709 fee and every and however many precincts 831 00:29:49,710 --> 00:29:51,659 he or she wants to get a recount. 832 00:29:51,660 --> 00:29:53,159 And then the State Board of Canvassers, 833 00:29:53,160 --> 00:29:54,839 which is the state committee that kind of 834 00:29:54,840 --> 00:29:57,809 decides that runs elections, 835 00:29:57,810 --> 00:29:59,670 decides how the recount will be done. 836 00:30:01,260 --> 00:30:03,209 And finally, in Pennsylvania, the margin 837 00:30:03,210 --> 00:30:04,980 was a bit bigger 44000 votes. 838 00:30:06,540 --> 00:30:07,709 Unfortunately, in Pennsylvania, they 839 00:30:07,710 --> 00:30:10,409 predominantly use paperless diaries, so 840 00:30:10,410 --> 00:30:12,239 70 percent of their votes are cast 841 00:30:12,240 --> 00:30:13,529 without any kind of paper, record or 842 00:30:13,530 --> 00:30:14,530 backup at all. 843 00:30:15,570 --> 00:30:17,069 And the law there is very, very 844 00:30:17,070 --> 00:30:18,119 complicated. 845 00:30:18,120 --> 00:30:20,279 It says that three citizens from 846 00:30:20,280 --> 00:30:22,739 every precinct that wants to be recounted 847 00:30:22,740 --> 00:30:25,499 must post a bond, 848 00:30:25,500 --> 00:30:27,779 swear that there is fraud. 849 00:30:27,780 --> 00:30:29,849 And you know, that's how you have 850 00:30:29,850 --> 00:30:31,439 to get a recount done in Pennsylvania. 851 00:30:31,440 --> 00:30:33,659 And bear in mind, there are tens 852 00:30:33,660 --> 00:30:34,949 of thousands of precincts in 853 00:30:34,950 --> 00:30:37,199 Pennsylvania, not tens of thousands. 854 00:30:37,200 --> 00:30:38,459 There are nine thousand or so precincts 855 00:30:38,460 --> 00:30:40,169 in Pennsylvania, so that's twenty seven 856 00:30:40,170 --> 00:30:41,999 thousand people we need to to swear and 857 00:30:42,000 --> 00:30:43,319 pay and et cetera. 858 00:30:44,550 --> 00:30:45,809 Fortunately, Pennsylvania does have a 859 00:30:45,810 --> 00:30:47,279 clause that states that an automatic 860 00:30:47,280 --> 00:30:49,439 recount is started statewide if the 861 00:30:49,440 --> 00:30:51,389 margin is less than half a percent. 862 00:30:51,390 --> 00:30:52,739 But unfortunately, the margin was just 863 00:30:52,740 --> 00:30:53,740 shy of that. 864 00:30:54,960 --> 00:30:56,099 So why not other states? 865 00:30:56,100 --> 00:30:58,019 One of the criticisms in particular from 866 00:30:58,020 --> 00:30:59,879 our president elect, is, you know, you 867 00:30:59,880 --> 00:31:02,189 pick three states that I won that Clinton 868 00:31:02,190 --> 00:31:03,539 was expected to win. Why aren't you 869 00:31:03,540 --> 00:31:05,459 auditing states that that she won, that I 870 00:31:05,460 --> 00:31:06,809 should have won? 871 00:31:06,810 --> 00:31:09,029 Why not other states where Donald Trump 872 00:31:09,030 --> 00:31:11,969 allegedly has heard tales of fraud? 873 00:31:11,970 --> 00:31:13,859 And so the three he pointed out Virginia, 874 00:31:13,860 --> 00:31:15,839 New Hampshire, California, New Hampshire. 875 00:31:15,840 --> 00:31:17,639 The deadline to get a recount done passed 876 00:31:17,640 --> 00:31:18,899 six days after the election. 877 00:31:18,900 --> 00:31:20,249 And as Alex just mentioned, we didn't get 878 00:31:20,250 --> 00:31:22,469 recount started until over two weeks 879 00:31:22,470 --> 00:31:23,879 after the election. 880 00:31:23,880 --> 00:31:25,829 In Virginia, it's actually illegal to 881 00:31:25,830 --> 00:31:27,899 recount an election if the margin 882 00:31:27,900 --> 00:31:29,939 of victory was greater than one percent, 883 00:31:29,940 --> 00:31:31,709 Clinton won by almost five percent. 884 00:31:31,710 --> 00:31:33,449 And in California, the margin of victory 885 00:31:33,450 --> 00:31:34,769 was 30 percent. 886 00:31:34,770 --> 00:31:36,209 I'm pretty sure Trump wouldn't have won 887 00:31:36,210 --> 00:31:38,099 California even if there was hacking for 888 00:31:38,100 --> 00:31:39,100 him. 889 00:31:40,050 --> 00:31:42,449 And all of this is to say that Trump 890 00:31:42,450 --> 00:31:44,069 could have also initiated recounts. 891 00:31:44,070 --> 00:31:46,019 Every candidate is entitled to initiate 892 00:31:46,020 --> 00:31:47,219 recounts in the United States. 893 00:31:47,220 --> 00:31:49,199 So if he really cared about voter fraud, 894 00:31:49,200 --> 00:31:50,400 he would have initiated a recount. 895 00:31:51,540 --> 00:31:53,939 So we're finally ready to 896 00:31:53,940 --> 00:31:55,679 kick off the recounts, and I'll let Alex 897 00:31:55,680 --> 00:31:56,680 tell you about that. 898 00:32:04,350 --> 00:32:06,209 All right. Time to let the recounting 899 00:32:06,210 --> 00:32:08,549 begin, so the recounts kicked 900 00:32:08,550 --> 00:32:10,859 off the day after Thanksgiving, November 901 00:32:10,860 --> 00:32:12,959 25th, when Jill Stein filed a 902 00:32:12,960 --> 00:32:15,899 petition in the state of Wisconsin 903 00:32:15,900 --> 00:32:18,059 to begin a formal 904 00:32:18,060 --> 00:32:20,279 recount. And this is just two days 905 00:32:20,280 --> 00:32:22,229 after she announced the recount 906 00:32:22,230 --> 00:32:23,230 initiative. 907 00:32:25,020 --> 00:32:27,299 The other states would soon 908 00:32:27,300 --> 00:32:29,249 follow in days to come. 909 00:32:29,250 --> 00:32:30,809 But at this point, there were only about 910 00:32:30,810 --> 00:32:32,939 two and a half weeks before that December 911 00:32:32,940 --> 00:32:34,319 13th deadline. 912 00:32:34,320 --> 00:32:36,809 So it was an incredible scramble. 913 00:32:36,810 --> 00:32:39,019 President Trump, of course, is, 914 00:32:39,020 --> 00:32:40,739 as you might expect, was not a fan. 915 00:32:40,740 --> 00:32:42,839 And I say that having looked at 916 00:32:42,840 --> 00:32:44,639 elections and election security in 917 00:32:44,640 --> 00:32:46,769 countries around the world, and I 918 00:32:46,770 --> 00:32:48,779 can say that no one who has won an 919 00:32:48,780 --> 00:32:50,909 election ever wants to see the 920 00:32:50,910 --> 00:32:53,009 election technology that was used to vote 921 00:32:53,010 --> 00:32:54,989 them into office criticized. 922 00:32:54,990 --> 00:32:56,879 I'm sure the shoe would be on the other 923 00:32:56,880 --> 00:32:58,589 foot if Clinton had won. 924 00:32:58,590 --> 00:33:00,329 And Donald Trump was alleging that the 925 00:33:00,330 --> 00:33:02,219 election had been rigged. 926 00:33:02,220 --> 00:33:04,169 But we have what we have. 927 00:33:04,170 --> 00:33:06,299 So Trump alleges, among 928 00:33:06,300 --> 00:33:07,709 other things, that this is all just a 929 00:33:07,710 --> 00:33:09,809 scam to help the Green Party. 930 00:33:09,810 --> 00:33:12,479 He and his lawyers would continuously 931 00:33:12,480 --> 00:33:14,699 oppose us in legal battles 932 00:33:14,700 --> 00:33:17,099 in all three states in an attempt to stop 933 00:33:17,100 --> 00:33:18,989 the recounts from beginning and halt them 934 00:33:18,990 --> 00:33:19,990 once they started. 935 00:33:21,360 --> 00:33:23,639 Now, Matt and I at this point 936 00:33:23,640 --> 00:33:25,739 basically just played a supporting 937 00:33:25,740 --> 00:33:27,959 role while others did the real work, 938 00:33:27,960 --> 00:33:30,059 including an army of Green Party 939 00:33:30,060 --> 00:33:32,909 volunteers. More than 10000 volunteers 940 00:33:32,910 --> 00:33:35,249 who worked at individual recount centers 941 00:33:35,250 --> 00:33:37,349 across the the states to 942 00:33:37,350 --> 00:33:39,749 observe who provided legal assistance 943 00:33:39,750 --> 00:33:41,909 on the ground, who provided 944 00:33:41,910 --> 00:33:43,709 assistance in filing petitions in 945 00:33:43,710 --> 00:33:46,199 Pennsylvania and also 946 00:33:46,200 --> 00:33:48,269 and critically importantly, we had 947 00:33:48,270 --> 00:33:50,279 excellent lawyers. 948 00:33:50,280 --> 00:33:51,839 We had assistance from the firm of 949 00:33:51,840 --> 00:33:53,939 Admiral Emory Celli, Brinckerhoff, 950 00:33:53,940 --> 00:33:56,369 Brinkerhoff, Brinkerhoff, excuse 951 00:33:56,370 --> 00:33:58,619 me, Matt and Abdi, 952 00:33:58,620 --> 00:34:00,989 who are specialists 953 00:34:00,990 --> 00:34:03,359 in litigation for topics 954 00:34:03,360 --> 00:34:05,549 like civil rights and election law. 955 00:34:06,570 --> 00:34:08,759 About half of the firm seems to have been 956 00:34:08,760 --> 00:34:10,408 involved in the recount efforts in one 957 00:34:10,409 --> 00:34:12,509 way or another. And over the course of 958 00:34:12,510 --> 00:34:14,609 about a week and a half, they 959 00:34:14,610 --> 00:34:16,829 initiated more than a half 960 00:34:16,830 --> 00:34:19,379 dozen different court actions 961 00:34:19,380 --> 00:34:21,448 in three states in state, 962 00:34:21,449 --> 00:34:23,519 federal and federal court in 963 00:34:23,520 --> 00:34:25,948 support of these recount efforts. 964 00:34:25,949 --> 00:34:28,169 We also got expert testimony from 965 00:34:28,170 --> 00:34:30,388 an army of leading electronic voting 966 00:34:30,389 --> 00:34:32,519 experts who rose to 967 00:34:32,520 --> 00:34:33,520 the occasion. 968 00:34:34,929 --> 00:34:36,730 So what does a recount look like? 969 00:34:39,949 --> 00:34:41,419 Well, it's actually not all that 970 00:34:41,420 --> 00:34:43,519 exciting, it's a bunch of people sitting 971 00:34:43,520 --> 00:34:45,678 around at card tables looking 972 00:34:45,679 --> 00:34:48,109 at stacks of paper mache, 973 00:34:48,110 --> 00:34:50,569 and I went and acted as recount 974 00:34:50,570 --> 00:34:52,789 observers in Washtenaw County, Michigan, 975 00:34:52,790 --> 00:34:54,979 where the university is and actually 976 00:34:54,980 --> 00:34:56,809 got to be a part of the process for a 977 00:34:56,810 --> 00:34:58,129 little while. 978 00:34:58,130 --> 00:35:00,379 So basically, what happens is 979 00:35:00,380 --> 00:35:02,689 that each table they recount one precinct 980 00:35:02,690 --> 00:35:03,679 at a time. 981 00:35:03,680 --> 00:35:05,959 You have two people who are employees 982 00:35:05,960 --> 00:35:08,479 of the county who look at each ballot, 983 00:35:08,480 --> 00:35:10,519 they count the ballots. 984 00:35:10,520 --> 00:35:12,049 Then they divide them into stacks, 985 00:35:12,050 --> 00:35:14,179 according to who the mark candidate was. 986 00:35:14,180 --> 00:35:15,979 Then they count the stacks. 987 00:35:15,980 --> 00:35:17,989 All the while, there were observers from 988 00:35:17,990 --> 00:35:20,209 any political candidate on the ballot 989 00:35:20,210 --> 00:35:22,399 who wants to have observers present 990 00:35:22,400 --> 00:35:23,719 who are watching to make sure 991 00:35:23,720 --> 00:35:25,609 everything's right and they can raise a 992 00:35:25,610 --> 00:35:27,699 challenge if they see any kind of error. 993 00:35:29,280 --> 00:35:31,409 So that's it, it's a laborious process it 994 00:35:31,410 --> 00:35:33,809 takes it took about three to four hours 995 00:35:33,810 --> 00:35:36,539 to recount a precinct while we watched 996 00:35:36,540 --> 00:35:38,789 of about 2000 ballots. 997 00:35:39,900 --> 00:35:42,149 But with parallel resolution, 998 00:35:42,150 --> 00:35:43,739 you can see there are a lot of tables 999 00:35:43,740 --> 00:35:45,209 there. This is the end of the day, so 1000 00:35:45,210 --> 00:35:47,249 most people have finished and gone home. 1001 00:35:47,250 --> 00:35:48,899 You can actually recount this process 1002 00:35:48,900 --> 00:35:50,039 fairly quickly. 1003 00:35:50,040 --> 00:35:52,139 It's nothing like the Florida 1004 00:35:52,140 --> 00:35:54,239 2000 recount, which was this 1005 00:35:54,240 --> 00:35:56,429 awful ancient punch card 1006 00:35:56,430 --> 00:35:58,769 technology and lasted for weeks 1007 00:35:58,770 --> 00:36:00,300 and caused endless litigation. 1008 00:36:01,750 --> 00:36:03,489 But a recount also looks like this 1009 00:36:03,490 --> 00:36:04,599 endless litigation. 1010 00:36:05,620 --> 00:36:07,809 Well, unfortunately, in all three 1011 00:36:07,810 --> 00:36:09,939 states, as I said, we had opposition 1012 00:36:09,940 --> 00:36:12,519 from either the Republican 1013 00:36:12,520 --> 00:36:15,129 Party supporting the Trump campaign 1014 00:36:15,130 --> 00:36:16,749 or in some cases from the state 1015 00:36:16,750 --> 00:36:18,369 governments themselves. 1016 00:36:18,370 --> 00:36:21,069 And we had to go 1017 00:36:21,070 --> 00:36:23,499 to federal court, to federal appeals 1018 00:36:23,500 --> 00:36:25,659 court, to in Michigan, to 1019 00:36:25,660 --> 00:36:27,759 the state Supreme Court in an 1020 00:36:27,760 --> 00:36:29,949 effort to try to have the recounts 1021 00:36:29,950 --> 00:36:30,950 proceed. 1022 00:36:32,120 --> 00:36:34,159 Recounts also look like this this is in 1023 00:36:34,160 --> 00:36:36,799 Michigan, where Donald Trump's lawyers 1024 00:36:36,800 --> 00:36:39,139 succeeded eventually in halting 1025 00:36:39,140 --> 00:36:41,599 the recount before it was completed. 1026 00:36:41,600 --> 00:36:43,549 This is of Green Party protest in front 1027 00:36:43,550 --> 00:36:45,739 of the state Supreme Court, asking 1028 00:36:45,740 --> 00:36:48,619 them to rule in our favor. 1029 00:36:48,620 --> 00:36:50,269 Recounts also look like this. 1030 00:36:50,270 --> 00:36:52,339 As I mentioned, it was a media circus. 1031 00:36:52,340 --> 00:36:54,709 This was the news just about every day. 1032 00:36:54,710 --> 00:36:57,079 I at one point, it's just so ridiculous. 1033 00:36:57,080 --> 00:36:59,149 Had a camera crew, 1034 00:36:59,150 --> 00:37:01,489 a photographer, a writer, a lighting 1035 00:37:01,490 --> 00:37:03,919 guy, another photographer, a sound person 1036 00:37:03,920 --> 00:37:06,019 and a director following me around 1037 00:37:06,020 --> 00:37:07,939 in an elevator as I was going from court 1038 00:37:07,940 --> 00:37:08,940 to court. 1039 00:37:10,570 --> 00:37:11,679 And it looks like this. 1040 00:37:11,680 --> 00:37:13,749 This is Jill Stein in Philadelphia, where 1041 00:37:13,750 --> 00:37:16,899 in Philadelphia we went to 1042 00:37:16,900 --> 00:37:19,179 the federal court to try to 1043 00:37:19,180 --> 00:37:21,339 convince a judge to give 1044 00:37:21,340 --> 00:37:23,649 us a recount of the entire 1045 00:37:23,650 --> 00:37:25,360 state or at least of select 1046 00:37:26,890 --> 00:37:29,199 select counties on constitutional 1047 00:37:29,200 --> 00:37:31,119 grounds because the state law didn't 1048 00:37:31,120 --> 00:37:33,129 provide any mechanism that we could use 1049 00:37:33,130 --> 00:37:34,630 to directly petition for it. 1050 00:37:37,340 --> 00:37:39,739 So recounts also look like this 1051 00:37:39,740 --> 00:37:41,929 camaraderie, people doing what 1052 00:37:41,930 --> 00:37:42,949 they think is right. 1053 00:37:42,950 --> 00:37:45,409 Many, many people working together, 1054 00:37:46,460 --> 00:37:49,309 but ultimately the recounts ended 1055 00:37:49,310 --> 00:37:51,259 and that's going to tell you what the 1056 00:37:51,260 --> 00:37:52,819 results were when the election was 1057 00:37:52,820 --> 00:37:53,820 decided. 1058 00:37:57,990 --> 00:38:00,089 So as you know, and as we've 1059 00:38:00,090 --> 00:38:02,129 sort of already discussed, Donald Trump 1060 00:38:02,130 --> 00:38:04,289 won the Electoral College on December 1061 00:38:04,290 --> 00:38:06,899 19th, submitted its votes. 1062 00:38:06,900 --> 00:38:09,249 And you know what's more to say, 1063 00:38:09,250 --> 00:38:10,250 right? 1064 00:38:11,980 --> 00:38:14,259 No state actually finished 1065 00:38:14,260 --> 00:38:16,149 a full hand recount when all was said and 1066 00:38:16,150 --> 00:38:17,049 done. 1067 00:38:17,050 --> 00:38:18,489 Wisconsin. 1068 00:38:18,490 --> 00:38:19,989 Almost all of their counties were hand 1069 00:38:19,990 --> 00:38:20,979 recount, but as I mentioned, their 1070 00:38:20,980 --> 00:38:23,109 counties could decide whether to do 1071 00:38:23,110 --> 00:38:24,849 hand or machine recount. 1072 00:38:24,850 --> 00:38:27,159 And you know, all told, about 1073 00:38:27,160 --> 00:38:28,209 400 votes changed 1074 00:38:29,410 --> 00:38:30,999 in Michigan. The recount ended after 1075 00:38:31,000 --> 00:38:33,279 about three days, again 1076 00:38:33,280 --> 00:38:35,589 due to opposition and 1077 00:38:35,590 --> 00:38:37,510 legal arguments over the term aggrieved. 1078 00:38:39,280 --> 00:38:40,899 So what did we what did we learn? 1079 00:38:40,900 --> 00:38:43,299 Or you know, what are the takeaways 1080 00:38:43,300 --> 00:38:45,699 here? Wisconsin was 1081 00:38:45,700 --> 00:38:47,859 a statewide recount. It did complete 51 1082 00:38:47,860 --> 00:38:50,769 counties counted by hand and then 21 1083 00:38:50,770 --> 00:38:52,749 didn't more or less. 1084 00:38:52,750 --> 00:38:54,549 There were eleven thousand eight hundred 1085 00:38:54,550 --> 00:38:56,409 eighty three ballot corrections found in 1086 00:38:56,410 --> 00:38:57,939 the recount, and this was things like 1087 00:38:57,940 --> 00:39:00,219 poll workers finding stacks of 20 1088 00:39:00,220 --> 00:39:02,319 or 30 ballots left over 1089 00:39:02,320 --> 00:39:03,549 that they forgot to count on election 1090 00:39:03,550 --> 00:39:04,449 night. 1091 00:39:04,450 --> 00:39:06,579 And by the way, that's more than half of 1092 00:39:06,580 --> 00:39:08,079 the margin. So had every one of those 1093 00:39:08,080 --> 00:39:10,179 corrections been for Hillary Clinton, the 1094 00:39:10,180 --> 00:39:12,369 results in Wisconsin would have flipped. 1095 00:39:12,370 --> 00:39:14,319 But as I mentioned, they weren't. 1096 00:39:14,320 --> 00:39:17,259 Only about 397 votes changed place. 1097 00:39:17,260 --> 00:39:18,789 And all told, we didn't find any 1098 00:39:18,790 --> 00:39:20,859 significant evidence of attack 1099 00:39:20,860 --> 00:39:22,059 in Michigan. 1100 00:39:22,060 --> 00:39:23,560 Like I said, it halted 1101 00:39:24,790 --> 00:39:26,859 10 counties actually finished recounting 1102 00:39:26,860 --> 00:39:28,359 12 other counties started, but didn't 1103 00:39:28,360 --> 00:39:29,289 finish. 1104 00:39:29,290 --> 00:39:31,749 And there are 83 counties in Michigan. 1105 00:39:31,750 --> 00:39:33,609 Over two million ballots were recounted, 1106 00:39:33,610 --> 00:39:34,959 which is forty three percent of the 1107 00:39:34,960 --> 00:39:37,089 original cast ballots. 1108 00:39:37,090 --> 00:39:39,369 We're still waiting for full data. 1109 00:39:39,370 --> 00:39:40,899 There was a net change of six hundred 1110 00:39:40,900 --> 00:39:42,369 fifty one votes. 1111 00:39:42,370 --> 00:39:45,009 That may change as more results come in. 1112 00:39:45,010 --> 00:39:46,809 But in general, no evidence of an attack 1113 00:39:46,810 --> 00:39:48,069 there, either. 1114 00:39:48,070 --> 00:39:49,869 In Pennsylvania, the recount never really 1115 00:39:49,870 --> 00:39:51,439 got off the ground. 1116 00:39:51,440 --> 00:39:52,780 I was defeated in federal court. 1117 00:39:53,830 --> 00:39:55,959 Only one county actually even started 1118 00:39:55,960 --> 00:39:58,419 a recount. As far as I know, 1119 00:39:58,420 --> 00:40:00,609 out of 67, they recounted by hand, and 1120 00:40:00,610 --> 00:40:02,769 they only recounted 143 of their 1121 00:40:02,770 --> 00:40:04,659 228 precincts. 1122 00:40:04,660 --> 00:40:06,219 They have not published results, but 1123 00:40:06,220 --> 00:40:07,809 presumably there wasn't an attack because 1124 00:40:07,810 --> 00:40:09,279 we didn't hear about it. 1125 00:40:09,280 --> 00:40:11,439 Hopefully, the lesson in Pennsylvania 1126 00:40:11,440 --> 00:40:12,879 was that it's really hard to get a 1127 00:40:12,880 --> 00:40:15,459 recount done in Pennsylvania. 1128 00:40:15,460 --> 00:40:17,529 Oh, other things we 1129 00:40:17,530 --> 00:40:19,989 learned in Detroit, Wayne 1130 00:40:19,990 --> 00:40:22,179 County, in Michigan, 1131 00:40:22,180 --> 00:40:24,280 there were huge systemic issues. 1132 00:40:25,540 --> 00:40:27,099 In 37 percent of precincts. 1133 00:40:27,100 --> 00:40:29,199 They found more votes for president than 1134 00:40:29,200 --> 00:40:31,389 there were actual ballots cast, 1135 00:40:31,390 --> 00:40:33,399 which is kind of a head scratcher. 1136 00:40:33,400 --> 00:40:35,499 And nearly half of the 1137 00:40:35,500 --> 00:40:37,659 ballots in Detroit were not eligible for 1138 00:40:37,660 --> 00:40:38,739 a recount. 1139 00:40:38,740 --> 00:40:39,740 With this, basically, 1140 00:40:40,810 --> 00:40:43,089 there's this notion of accountability 1141 00:40:43,090 --> 00:40:45,459 where in order to recount the ballot, 1142 00:40:45,460 --> 00:40:47,019 poll workers have to show that the chain 1143 00:40:47,020 --> 00:40:48,609 of custody from the end of the election 1144 00:40:48,610 --> 00:40:49,779 until when the ballots are being 1145 00:40:49,780 --> 00:40:51,219 recounted was upheld. 1146 00:40:51,220 --> 00:40:52,779 And this includes things like poll book 1147 00:40:52,780 --> 00:40:55,059 logs, tamper proof seals 1148 00:40:55,060 --> 00:40:57,189 on on the bags that have ballots in them 1149 00:40:57,190 --> 00:41:00,009 and so on. So if you're an attacker, 1150 00:41:00,010 --> 00:41:01,419 all you have to do, you know, let's say 1151 00:41:01,420 --> 00:41:02,889 you hack a machine and you want to hide 1152 00:41:02,890 --> 00:41:04,509 your tracks. All you have to do is go 1153 00:41:04,510 --> 00:41:06,549 pull a few seals off of these bags or 1154 00:41:06,550 --> 00:41:08,319 tear open the ballot bags. 1155 00:41:08,320 --> 00:41:10,509 And you know, no one will ever find 1156 00:41:10,510 --> 00:41:12,399 out that you hacked anything. 1157 00:41:12,400 --> 00:41:14,019 So even when you find evidence that maybe 1158 00:41:14,020 --> 00:41:15,699 something went wrong, nobody looks at 1159 00:41:17,530 --> 00:41:19,059 using the available evidence that we did 1160 00:41:19,060 --> 00:41:20,859 get from the recounts. 1161 00:41:20,860 --> 00:41:23,229 We tried to sort of extract information 1162 00:41:23,230 --> 00:41:24,759 and draw conclusions. 1163 00:41:24,760 --> 00:41:26,229 What we would really like to do is what's 1164 00:41:26,230 --> 00:41:28,299 called a statistical risk limiting 1165 00:41:28,300 --> 00:41:29,300 audit. 1166 00:41:30,220 --> 00:41:31,659 Or basically you handcuffed a very small 1167 00:41:31,660 --> 00:41:34,089 sample of ballots randomly chosen 1168 00:41:34,090 --> 00:41:35,619 across a state or across the county. 1169 00:41:35,620 --> 00:41:37,150 Or however, you would like to do it 1170 00:41:38,260 --> 00:41:40,239 and you hand count them, and you can show 1171 00:41:40,240 --> 00:41:41,919 statistically that the hand counted 1172 00:41:41,920 --> 00:41:44,199 results verifies these statewide 1173 00:41:44,200 --> 00:41:46,419 results. This is significantly cheaper 1174 00:41:46,420 --> 00:41:48,249 than a full state recount. 1175 00:41:48,250 --> 00:41:50,379 But unfortunately, most law state laws 1176 00:41:50,380 --> 00:41:52,089 and in fact, I think no state actually 1177 00:41:52,090 --> 00:41:54,309 allows risk limiting audits, except maybe 1178 00:41:54,310 --> 00:41:55,310 one. 1179 00:41:55,660 --> 00:41:56,859 What we actually wound up with, of 1180 00:41:56,860 --> 00:41:58,629 course, was incomplete non random county 1181 00:41:58,630 --> 00:42:00,099 level samples. 1182 00:42:00,100 --> 00:42:01,009 So what do we do? 1183 00:42:01,010 --> 00:42:02,710 How do we draw conclusions from this? 1184 00:42:04,450 --> 00:42:06,549 Well, we can't rule out some attack 1185 00:42:06,550 --> 00:42:07,479 scenarios. 1186 00:42:07,480 --> 00:42:08,679 For instance, statewide fraud. 1187 00:42:08,680 --> 00:42:10,299 You know, it's unlikely in a recount that 1188 00:42:10,300 --> 00:42:11,679 every single ballot, you know, it was 1189 00:42:11,680 --> 00:42:13,809 hacked or that as if every ballot 1190 00:42:13,810 --> 00:42:15,039 and every piece or a ballot in every 1191 00:42:15,040 --> 00:42:16,149 precinct was hacked. 1192 00:42:17,440 --> 00:42:18,489 What about other scenarios? 1193 00:42:18,490 --> 00:42:20,139 So we're still waiting for a complete 1194 00:42:20,140 --> 00:42:22,239 analysis. But the idea is that we want to 1195 00:42:22,240 --> 00:42:24,939 test the likelihood of a probable attack 1196 00:42:24,940 --> 00:42:27,249 of catching a probable attack. 1197 00:42:27,250 --> 00:42:29,769 So our model is an attacker compromises 1198 00:42:29,770 --> 00:42:32,020 some subset of counties 1199 00:42:33,070 --> 00:42:34,989 at random. We presume that the attackers 1200 00:42:34,990 --> 00:42:36,459 won't have the ability to choose where 1201 00:42:36,460 --> 00:42:39,009 they get to attack, modulo 1202 00:42:39,010 --> 00:42:40,010 a few things. 1203 00:42:41,170 --> 00:42:44,139 And basically, we sample 1204 00:42:44,140 --> 00:42:45,939 counties at random and change 10 percent 1205 00:42:45,940 --> 00:42:48,189 of the votes until we have enough to flip 1206 00:42:48,190 --> 00:42:49,239 the election in that state 1207 00:42:50,740 --> 00:42:53,019 if the randomly selected counties 1208 00:42:53,020 --> 00:42:54,579 were recounted by hand. 1209 00:42:54,580 --> 00:42:56,260 Then we consider that attack detected. 1210 00:42:57,820 --> 00:42:59,739 So what we find is that in Wisconsin, we 1211 00:42:59,740 --> 00:43:01,359 ran a million simulations of this attack 1212 00:43:01,360 --> 00:43:02,979 in every state. And what we find is that 1213 00:43:02,980 --> 00:43:04,989 in Wisconsin, you only need on average 1214 00:43:04,990 --> 00:43:07,329 seven counties to throw the result. 1215 00:43:08,480 --> 00:43:10,989 And fortunately, because Wisconsin had 1216 00:43:10,990 --> 00:43:12,759 almost all. States, too, are all counties 1217 00:43:12,760 --> 00:43:14,079 do a hand recount. 1218 00:43:14,080 --> 00:43:15,549 The chances of being undetected are 1219 00:43:15,550 --> 00:43:16,929 pretty, pretty low. 1220 00:43:16,930 --> 00:43:18,549 You know, you have a five percent success 1221 00:43:18,550 --> 00:43:20,259 rate there. 1222 00:43:20,260 --> 00:43:21,759 In Michigan, there was a much tighter 1223 00:43:21,760 --> 00:43:24,099 margin of, you know, only 11000 votes. 1224 00:43:25,570 --> 00:43:27,819 So you need an attacker needs very 1225 00:43:27,820 --> 00:43:29,919 much fewer counties to flip to 1226 00:43:29,920 --> 00:43:32,379 rig the results for. 1227 00:43:32,380 --> 00:43:33,909 And because the Michigan recount didn't 1228 00:43:33,910 --> 00:43:36,159 actually finish the 1229 00:43:36,160 --> 00:43:38,079 chance of the attack, going undetected is 1230 00:43:38,080 --> 00:43:39,669 is higher than it was in Wisconsin. 1231 00:43:39,670 --> 00:43:41,109 Albeit it's still not that high, 1232 00:43:41,110 --> 00:43:43,329 considering only a small 1233 00:43:43,330 --> 00:43:44,559 fraction of counties were actually 1234 00:43:44,560 --> 00:43:45,459 recounted by hand. 1235 00:43:45,460 --> 00:43:47,589 In Michigan, Pennsylvania, there was 1236 00:43:47,590 --> 00:43:49,989 a wider margin, so you need more 1237 00:43:49,990 --> 00:43:52,089 more counties to throw the results. 1238 00:43:52,090 --> 00:43:54,309 And as we mentioned, only one county, 1239 00:43:54,310 --> 00:43:55,899 only part of one county did a hand 1240 00:43:55,900 --> 00:43:57,369 recount. So the chance of going 1241 00:43:57,370 --> 00:43:58,899 undetected, there was a very high. 1242 00:44:01,020 --> 00:44:03,299 So what have we learned over all 1243 00:44:03,300 --> 00:44:05,639 the recounts support that 1244 00:44:05,640 --> 00:44:08,079 the election outcome was correct, that 1245 00:44:08,080 --> 00:44:10,349 we can I can sleep at night knowing 1246 00:44:10,350 --> 00:44:11,849 that Donald Trump won the election and 1247 00:44:11,850 --> 00:44:14,039 having no doubt there 1248 00:44:14,040 --> 00:44:16,259 is not strong evidence that there 1249 00:44:16,260 --> 00:44:18,479 wasn't fraud or that there wasn't a cyber 1250 00:44:18,480 --> 00:44:20,609 attack, though, because recounts were 1251 00:44:20,610 --> 00:44:21,869 incomplete because they're sort of an 1252 00:44:21,870 --> 00:44:23,519 imperfect way to do this, you know, to 1253 00:44:23,520 --> 00:44:24,520 verify. 1254 00:44:25,170 --> 00:44:26,170 So, 1255 00:44:27,690 --> 00:44:29,609 yeah, like we said, cyber attacks already 1256 00:44:29,610 --> 00:44:30,690 happened before 1257 00:44:31,800 --> 00:44:34,199 we did get a congressional investigation 1258 00:44:34,200 --> 00:44:35,789 out of this. President Obama has ordered 1259 00:44:35,790 --> 00:44:37,619 an investigation before he leaves office. 1260 00:44:38,760 --> 00:44:41,219 And, you know, we got 1261 00:44:41,220 --> 00:44:43,319 intelligence agencies to agree that, 1262 00:44:43,320 --> 00:44:44,609 you know, there was hacking that did 1263 00:44:44,610 --> 00:44:45,989 happen that probably influence the 1264 00:44:45,990 --> 00:44:47,009 outcome. 1265 00:44:47,010 --> 00:44:48,010 But overall, 1266 00:44:49,500 --> 00:44:50,849 we can sleep at night, but we can't 1267 00:44:50,850 --> 00:44:51,929 completely rule out fraud. 1268 00:45:01,190 --> 00:45:04,099 So the recounts didn't complete 1269 00:45:04,100 --> 00:45:05,599 what I thought at the beginning, as I 1270 00:45:05,600 --> 00:45:08,149 said in my medium piece, was that, well, 1271 00:45:08,150 --> 00:45:10,999 the election probably wasn't hacked. 1272 00:45:11,000 --> 00:45:12,739 What did we learn from the recounts, what 1273 00:45:12,740 --> 00:45:14,989 we learned that it even more 1274 00:45:14,990 --> 00:45:16,489 probably wasn't hacked? 1275 00:45:16,490 --> 00:45:19,159 We get additional basis for confidence 1276 00:45:19,160 --> 00:45:20,160 in the election. 1277 00:45:21,170 --> 00:45:23,299 But at the same time, along 1278 00:45:23,300 --> 00:45:25,759 the way, we found that hacking 1279 00:45:25,760 --> 00:45:27,259 an election in the U.S. 1280 00:45:27,260 --> 00:45:29,479 for president would be, well, even 1281 00:45:29,480 --> 00:45:31,639 easier than I thought when I began 1282 00:45:31,640 --> 00:45:33,289 this process. 1283 00:45:33,290 --> 00:45:34,849 We have the vulnerable machines. 1284 00:45:34,850 --> 00:45:37,159 Of course we knew that before, but how 1285 00:45:37,160 --> 00:45:39,349 central the points of attack are 1286 00:45:39,350 --> 00:45:41,449 these small companies that control voting 1287 00:45:41,450 --> 00:45:43,549 machines across many states 1288 00:45:43,550 --> 00:45:44,989 and are such small businesses? 1289 00:45:44,990 --> 00:45:47,329 Well, that was news to me and many 1290 00:45:47,330 --> 00:45:48,330 other people. 1291 00:45:49,130 --> 00:45:51,559 Also shocking is how unlikely 1292 00:45:51,560 --> 00:45:53,629 states are to look at any of 1293 00:45:53,630 --> 00:45:56,089 the paper, even in a surprising and close 1294 00:45:56,090 --> 00:45:57,589 election like this. 1295 00:45:57,590 --> 00:45:59,869 Absent a recount, probably none 1296 00:45:59,870 --> 00:46:01,729 of the paper would have been looked at. 1297 00:46:01,730 --> 00:46:03,559 But with the recounts, at least we had 1298 00:46:03,560 --> 00:46:05,479 some enough to gain some amount of 1299 00:46:05,480 --> 00:46:06,830 statistical confidence, 1300 00:46:08,450 --> 00:46:10,819 even if a candidate can force a recount, 1301 00:46:10,820 --> 00:46:12,529 and this is probably the most damning 1302 00:46:12,530 --> 00:46:15,349 thing about the entire experience. 1303 00:46:15,350 --> 00:46:17,479 Well, there are opportunities for the 1304 00:46:17,480 --> 00:46:19,669 apparent winner to try to stop 1305 00:46:19,670 --> 00:46:21,439 it. There are many, many opportunities 1306 00:46:21,440 --> 00:46:23,509 and in fact, they're quite likely to be 1307 00:46:23,510 --> 00:46:25,249 successful. 1308 00:46:25,250 --> 00:46:27,409 Whoever wins is likely to 1309 00:46:27,410 --> 00:46:30,499 do this Republican or Democrat. 1310 00:46:30,500 --> 00:46:32,899 So unfortunately, although we have 1311 00:46:32,900 --> 00:46:34,609 some amount of physical evidence we could 1312 00:46:34,610 --> 00:46:36,709 go and look at, we probably will never 1313 00:46:36,710 --> 00:46:38,449 get the chance. 1314 00:46:38,450 --> 00:46:40,669 What we need in the United States quite 1315 00:46:40,670 --> 00:46:43,219 badly at this point is some specific 1316 00:46:43,220 --> 00:46:45,679 reform to the election process. 1317 00:46:45,680 --> 00:46:48,649 We need this because even if the 2016 1318 00:46:48,650 --> 00:46:51,049 election wasn't hacked, the 2020 1319 00:46:51,050 --> 00:46:53,029 election might well be. 1320 00:46:53,030 --> 00:46:55,309 We're facing increasingly powerful 1321 00:46:55,310 --> 00:46:57,769 and aggressive state level attackers 1322 00:46:57,770 --> 00:46:59,989 in every country, including 1323 00:46:59,990 --> 00:47:02,659 attackers targeting the US. 1324 00:47:02,660 --> 00:47:04,909 We need some effective defenses 1325 00:47:04,910 --> 00:47:06,649 to keep them from undermining our core 1326 00:47:06,650 --> 00:47:08,329 democracy. 1327 00:47:08,330 --> 00:47:10,729 First, it's common sense 1328 00:47:10,730 --> 00:47:12,259 we should be hardening our voting 1329 00:47:12,260 --> 00:47:14,359 technology to make it more resistant 1330 00:47:14,360 --> 00:47:16,759 to attacks, to buffer overflows to basic 1331 00:47:16,760 --> 00:47:19,069 security vulnerabilities. 1332 00:47:19,070 --> 00:47:21,799 But even this, it's not sufficient. 1333 00:47:21,800 --> 00:47:24,319 This raises the bar to an attack, 1334 00:47:24,320 --> 00:47:26,029 but we need to make sure that we're 1335 00:47:26,030 --> 00:47:28,339 getting a physical safeguard 1336 00:47:28,340 --> 00:47:30,799 a common sense kind of security 1337 00:47:30,800 --> 00:47:32,479 audit of backup. 1338 00:47:32,480 --> 00:47:34,519 By having a paper record in place for 1339 00:47:34,520 --> 00:47:36,589 every vote, we have to go from 1340 00:47:36,590 --> 00:47:38,899 70 percent to 100 percent. 1341 00:47:38,900 --> 00:47:40,549 And there's still a number of states that 1342 00:47:40,550 --> 00:47:42,949 need to take action to do so. 1343 00:47:42,950 --> 00:47:45,169 But most importantly, we need 1344 00:47:45,170 --> 00:47:47,029 to make sure that every state actually 1345 00:47:47,030 --> 00:47:49,549 makes use of that evidence 1346 00:47:49,550 --> 00:47:50,809 and to do that. 1347 00:47:50,810 --> 00:47:53,119 We need not recounts, recounts 1348 00:47:53,120 --> 00:47:54,289 or a hack. 1349 00:47:54,290 --> 00:47:56,389 We need statistical risk limiting 1350 00:47:56,390 --> 00:47:58,789 audit to take place after every major 1351 00:47:58,790 --> 00:48:00,019 election. 1352 00:48:00,020 --> 00:48:02,149 These are a cheap and easy to 1353 00:48:02,150 --> 00:48:04,369 use mechanism that states can use 1354 00:48:04,370 --> 00:48:06,469 to make sure the electronic totals are 1355 00:48:06,470 --> 00:48:09,259 right by looking at enough of the paper. 1356 00:48:09,260 --> 00:48:11,389 States need to act now 1357 00:48:11,390 --> 00:48:13,069 to make sure that these are going to be 1358 00:48:13,070 --> 00:48:15,349 in place. All of these reforms in time 1359 00:48:15,350 --> 00:48:17,809 for 2018 and 2020, 1360 00:48:17,810 --> 00:48:20,119 the next presidential election. 1361 00:48:20,120 --> 00:48:22,489 But it's going to be an uphill battle. 1362 00:48:22,490 --> 00:48:25,399 Unfortunately, because of our distributed 1363 00:48:25,400 --> 00:48:27,589 federal system, states 1364 00:48:27,590 --> 00:48:29,509 individually are likely going to need to 1365 00:48:29,510 --> 00:48:31,609 act. I think we're unlikely to 1366 00:48:31,610 --> 00:48:33,649 see federal legislation to improve 1367 00:48:33,650 --> 00:48:35,869 electoral integrity under the coming 1368 00:48:35,870 --> 00:48:37,789 Trump administration. 1369 00:48:37,790 --> 00:48:39,319 So thank you very much. 1370 00:48:39,320 --> 00:48:40,760 We have some time for questions. 1371 00:48:45,470 --> 00:48:47,479 OK, perfect. 1372 00:48:47,480 --> 00:48:48,620 We have about. 1373 00:48:55,790 --> 00:48:57,439 All right. Thanks a lot, Matt and Alex, 1374 00:48:57,440 --> 00:48:59,149 we have about 10 minutes for questions, 1375 00:48:59,150 --> 00:49:01,069 so if you have a question, please line up 1376 00:49:01,070 --> 00:49:03,199 at one of the microphones in 1377 00:49:03,200 --> 00:49:04,519 the room. 1378 00:49:04,520 --> 00:49:06,619 And I think we'll just start 1379 00:49:06,620 --> 00:49:07,969 off with microphone one, please. 1380 00:49:10,010 --> 00:49:12,179 We do not believe in people 1381 00:49:12,180 --> 00:49:14,349 voting so much that 1382 00:49:14,350 --> 00:49:16,459 way because I lived 1383 00:49:16,460 --> 00:49:19,699 half of my life during Soviet occupation 1384 00:49:19,700 --> 00:49:22,279 and I have seen many 1385 00:49:22,280 --> 00:49:24,799 people vote things that ended. 1386 00:49:24,800 --> 00:49:27,079 Ninety nine point nine 1387 00:49:27,080 --> 00:49:29,150 percent for Communist Party. 1388 00:49:30,380 --> 00:49:32,629 My question is 1389 00:49:32,630 --> 00:49:34,999 related to the financing and then 1390 00:49:35,000 --> 00:49:36,129 methods of press. 1391 00:49:36,130 --> 00:49:38,209 So how to make a government to 1392 00:49:38,210 --> 00:49:40,010 make elections better? 1393 00:49:41,330 --> 00:49:43,489 That's a very, very big thing that 1394 00:49:43,490 --> 00:49:45,649 I had to miss your previous 1395 00:49:45,650 --> 00:49:47,959 talk here about 1396 00:49:47,960 --> 00:49:50,569 Estonian voting. 1397 00:49:50,570 --> 00:49:52,819 So I remember what 1398 00:49:52,820 --> 00:49:53,839 it was. 1399 00:49:53,840 --> 00:49:55,820 Professor Alderman's team 1400 00:49:56,990 --> 00:49:59,449 was invited into Estonia 1401 00:49:59,450 --> 00:50:01,969 by a political party, which 1402 00:50:01,970 --> 00:50:04,429 later was seen, 1403 00:50:04,430 --> 00:50:07,339 at least asking money from Putin. 1404 00:50:07,340 --> 00:50:10,079 I don't know whether they got or not. 1405 00:50:10,080 --> 00:50:11,089 So could you 1406 00:50:11,090 --> 00:50:13,279 please come to be happy to answer that? 1407 00:50:13,280 --> 00:50:15,499 Yes, yes, yes, absolutely. 1408 00:50:15,500 --> 00:50:17,629 And then 1409 00:50:17,630 --> 00:50:20,150 we compare the definition of terrorism. 1410 00:50:21,680 --> 00:50:22,489 May I answer your 1411 00:50:22,490 --> 00:50:23,399 question, sir? 1412 00:50:23,400 --> 00:50:24,479 Yeah. 1413 00:50:24,480 --> 00:50:26,179 So let me let me address first. 1414 00:50:26,180 --> 00:50:28,309 The point about paper 1415 00:50:28,310 --> 00:50:30,709 is also vulnerable, and that's absolutely 1416 00:50:30,710 --> 00:50:32,839 true. We have a long history through all 1417 00:50:32,840 --> 00:50:34,969 use of paper and democracy of paper 1418 00:50:34,970 --> 00:50:36,769 also being hacked. 1419 00:50:36,770 --> 00:50:38,869 The key is that we have an opportunity 1420 00:50:38,870 --> 00:50:41,779 now because we're combining computers 1421 00:50:41,780 --> 00:50:44,059 and paper to get the best of both 1422 00:50:44,060 --> 00:50:46,349 worlds, a digital record of the vote and 1423 00:50:46,350 --> 00:50:48,589 the physical record of the vote, as long 1424 00:50:48,590 --> 00:50:50,659 as we check to make sure they agree. 1425 00:50:50,660 --> 00:50:52,789 If we don't check that they agree, then 1426 00:50:52,790 --> 00:50:54,679 we might as well just have the one or the 1427 00:50:54,680 --> 00:50:56,869 other. Also, I'm not sure that we 1428 00:50:56,870 --> 00:50:58,909 at the University of Michigan could hack 1429 00:50:58,910 --> 00:51:01,069 into all the paper ballots across 1430 00:51:01,070 --> 00:51:03,049 multiple states sufficient to change the 1431 00:51:03,050 --> 00:51:04,789 presidential election. 1432 00:51:04,790 --> 00:51:06,739 But I'm pretty sure my undergraduate 1433 00:51:06,740 --> 00:51:09,109 security course could have changed 1434 00:51:09,110 --> 00:51:10,939 the outcome of the presidential election 1435 00:51:10,940 --> 00:51:12,859 this year. It really is that bad. 1436 00:51:12,860 --> 00:51:13,519 Thank you. 1437 00:51:13,520 --> 00:51:15,979 How to make the voting 1438 00:51:15,980 --> 00:51:17,029 more transparent. 1439 00:51:17,030 --> 00:51:18,709 Sorry, can we please just have one 1440 00:51:18,710 --> 00:51:19,969 question per person? We have lots of 1441 00:51:19,970 --> 00:51:22,129 people who want to ask questions. 1442 00:51:22,130 --> 00:51:23,359 Happy to talk to you one on one 1443 00:51:23,360 --> 00:51:24,360 afterwards. 1444 00:51:24,860 --> 00:51:26,149 Just just a quick heads up. 1445 00:51:26,150 --> 00:51:28,549 If you really have to leave right now, 1446 00:51:28,550 --> 00:51:30,769 please do so as quietly as possible so we 1447 00:51:30,770 --> 00:51:33,229 can finish this Q&A. 1448 00:51:33,230 --> 00:51:35,299 And also, if you if you're asking a 1449 00:51:35,300 --> 00:51:37,129 question, please make it precise and 1450 00:51:37,130 --> 00:51:39,169 short so we can ask as many questions as 1451 00:51:39,170 --> 00:51:40,249 possible. Mike two Please. 1452 00:51:41,390 --> 00:51:43,699 I want to ask if it wouldn't 1453 00:51:43,700 --> 00:51:45,859 have been easier to get your hands on 1454 00:51:45,860 --> 00:51:48,049 on the hardware and the software and 1455 00:51:48,050 --> 00:51:50,329 just check for any vote 1456 00:51:50,330 --> 00:51:52,579 or if it has been compromised. 1457 00:51:52,580 --> 00:51:54,529 I mean, pay some guys off and get your 1458 00:51:54,530 --> 00:51:55,530 hands on it. 1459 00:51:56,540 --> 00:51:58,249 Well, we we did check. 1460 00:51:58,250 --> 00:52:00,679 We did try in Pennsylvania to get access 1461 00:52:00,680 --> 00:52:02,449 to the hardware, in fact, because that's 1462 00:52:02,450 --> 00:52:04,459 often the only record of the vote is the 1463 00:52:04,460 --> 00:52:06,199 one that's made in the computer's memory 1464 00:52:06,200 --> 00:52:07,969 in Pennsylvania. 1465 00:52:07,970 --> 00:52:09,619 That would have been very difficult to 1466 00:52:09,620 --> 00:52:11,719 complete a forensic analysis in 1467 00:52:11,720 --> 00:52:13,609 the time available and only a couple of 1468 00:52:13,610 --> 00:52:15,319 weeks before the deadline. 1469 00:52:15,320 --> 00:52:17,509 And it's not really as direct a 1470 00:52:17,510 --> 00:52:19,579 way of checking as looking at the paper 1471 00:52:19,580 --> 00:52:21,779 if you have access to paper records, too. 1472 00:52:21,780 --> 00:52:24,349 But absolutely, I support forensic 1473 00:52:24,350 --> 00:52:26,389 analysis of the voting equipment to I 1474 00:52:26,390 --> 00:52:28,189 think that's just another kind of 1475 00:52:28,190 --> 00:52:29,190 physical evidence. 1476 00:52:31,140 --> 00:52:32,189 OK, perfect. 1477 00:52:32,190 --> 00:52:33,899 Mike, for peace. 1478 00:52:33,900 --> 00:52:34,949 Thank you. 1479 00:52:34,950 --> 00:52:37,109 In the history of the world, 1480 00:52:37,110 --> 00:52:39,269 to the best of your knowledge, has it 1481 00:52:39,270 --> 00:52:41,429 ever been a case for the 1482 00:52:41,430 --> 00:52:43,829 election process was rebooted 1483 00:52:43,830 --> 00:52:46,679 because it was proven to be compromised? 1484 00:52:48,210 --> 00:52:49,739 There's a first time for everything. 1485 00:52:54,010 --> 00:52:56,109 All right. I think we're going to squeeze 1486 00:52:56,110 --> 00:52:58,869 in the question from Mike three, please. 1487 00:52:58,870 --> 00:53:01,599 OK. This risk limiting 1488 00:53:01,600 --> 00:53:03,159 hand counting approach. 1489 00:53:03,160 --> 00:53:05,169 Does it have some requirements that will 1490 00:53:05,170 --> 00:53:07,179 make it unfit for other countries than 1491 00:53:07,180 --> 00:53:08,180 the US? 1492 00:53:10,920 --> 00:53:11,969 Does it? 1493 00:53:11,970 --> 00:53:14,039 I'm sorry. That makes it fit for other so 1494 00:53:14,040 --> 00:53:17,159 this statistical 1495 00:53:17,160 --> 00:53:18,839 approach does it to have requirements 1496 00:53:18,840 --> 00:53:20,339 that are unique to the U.S., you 1497 00:53:20,340 --> 00:53:22,649 know, it's limiting on can be used 1498 00:53:22,650 --> 00:53:23,789 just about anywhere. There's paper to 1499 00:53:23,790 --> 00:53:26,189 count. And it can also be used. 1500 00:53:26,190 --> 00:53:27,929 It's not strictly to the first past the 1501 00:53:27,930 --> 00:53:29,489 post voting method that we use in the 1502 00:53:29,490 --> 00:53:30,490 states. It can be used for 1503 00:53:32,100 --> 00:53:33,929 any any of the methods that you know, 1504 00:53:33,930 --> 00:53:35,069 just about anywhere uses. 1505 00:53:35,070 --> 00:53:36,070 OK, thank you. 1506 00:53:36,780 --> 00:53:38,849 OK, I think we have a question from the 1507 00:53:38,850 --> 00:53:41,009 people watching the livestream 1508 00:53:41,010 --> 00:53:41,999 single angel, please. 1509 00:53:42,000 --> 00:53:43,619 Actually, we have to. 1510 00:53:43,620 --> 00:53:45,749 The first question is why is Senate 1511 00:53:45,750 --> 00:53:47,969 always done a recount? 1512 00:53:47,970 --> 00:53:48,970 Good question. 1513 00:53:50,310 --> 00:53:51,839 It is a really great question. 1514 00:53:51,840 --> 00:53:53,159 It should always be done. 1515 00:53:53,160 --> 00:53:55,319 Some kind of risk limiting audit 1516 00:53:55,320 --> 00:53:56,729 anyway, looking at enough of the 1517 00:53:56,730 --> 00:53:58,979 evidence. But in part, it's because 1518 00:53:58,980 --> 00:54:01,349 the law hasn't caught up to technology 1519 00:54:01,350 --> 00:54:03,569 right back when we were only voting 1520 00:54:03,570 --> 00:54:04,979 on paper. 1521 00:54:04,980 --> 00:54:06,089 There wasn't. 1522 00:54:06,090 --> 00:54:08,339 There was a hand count already of 1523 00:54:08,340 --> 00:54:10,499 every ballot before we had 1524 00:54:10,500 --> 00:54:11,789 any kind of voting machine. 1525 00:54:11,790 --> 00:54:13,829 But since the introduction of technology 1526 00:54:13,830 --> 00:54:16,139 and especially such hackable technology, 1527 00:54:16,140 --> 00:54:18,329 we need to update the law to make sure 1528 00:54:18,330 --> 00:54:20,369 some amount of physical auditing happens 1529 00:54:20,370 --> 00:54:21,370 every time. 1530 00:54:21,960 --> 00:54:24,389 And we have a second question which 1531 00:54:24,390 --> 00:54:26,459 asks, in your own opinion, would it 1532 00:54:26,460 --> 00:54:28,829 make sense to get sensible 1533 00:54:28,830 --> 00:54:30,060 loss in all the states? 1534 00:54:31,830 --> 00:54:33,479 To get pardon, can you repeat that? 1535 00:54:34,770 --> 00:54:36,089 Wouldn't it make sense in your own 1536 00:54:36,090 --> 00:54:38,369 opinion to get clean laws 1537 00:54:38,370 --> 00:54:40,589 in all the states, so 1538 00:54:40,590 --> 00:54:42,869 it's more sensible to understand 1539 00:54:42,870 --> 00:54:44,609 if there is a rigging or not? 1540 00:54:44,610 --> 00:54:46,529 Certainly. But one of the issues is the 1541 00:54:46,530 --> 00:54:48,119 United States is a federal system, so the 1542 00:54:48,120 --> 00:54:49,949 states ultimately have most of the rights 1543 00:54:49,950 --> 00:54:51,479 over what they can do. 1544 00:54:51,480 --> 00:54:52,949 And so there are things like the Federal 1545 00:54:52,950 --> 00:54:55,079 Election Commission that do try to to, 1546 00:54:55,080 --> 00:54:56,969 at a federal level, enforce policy. 1547 00:54:56,970 --> 00:54:59,159 But it's much more complicated and 1548 00:54:59,160 --> 00:55:01,349 difficult constitutionally even 1549 00:55:01,350 --> 00:55:03,329 than just saying all votes will be done 1550 00:55:03,330 --> 00:55:04,229 this way. 1551 00:55:04,230 --> 00:55:05,999 Believe it or not, in the United States, 1552 00:55:06,000 --> 00:55:08,159 the Constitution doesn't even say that 1553 00:55:08,160 --> 00:55:10,079 the states have to have an election to 1554 00:55:10,080 --> 00:55:12,149 decide who their presidential votes will 1555 00:55:12,150 --> 00:55:13,229 go to. 1556 00:55:13,230 --> 00:55:15,119 They can just pick it by whatever 1557 00:55:15,120 --> 00:55:17,819 legislate method the Legislature decides. 1558 00:55:17,820 --> 00:55:19,859 So we really do have to go state by 1559 00:55:19,860 --> 00:55:21,959 state. I'm glad they at least ask the 1560 00:55:21,960 --> 00:55:22,960 people what they want. 1561 00:55:25,470 --> 00:55:26,759 All right. 1562 00:55:26,760 --> 00:55:28,409 Mike five. Please. 1563 00:55:28,410 --> 00:55:30,509 Hi. Election reform 1564 00:55:30,510 --> 00:55:32,699 is itself a political process, so 1565 00:55:32,700 --> 00:55:34,829 I wanted to ask how your efforts 1566 00:55:34,830 --> 00:55:37,169 to improve the technical aspects 1567 00:55:37,170 --> 00:55:39,359 of actual voting interact 1568 00:55:39,360 --> 00:55:41,579 with the wider issues 1569 00:55:41,580 --> 00:55:43,109 that other people are struggling over, 1570 00:55:43,110 --> 00:55:45,509 such as voter suppression due to felony 1571 00:55:45,510 --> 00:55:47,819 convictions, positioning 1572 00:55:47,820 --> 00:55:49,859 of ballot boxes and all the rest that's 1573 00:55:49,860 --> 00:55:51,659 well known in the US. 1574 00:55:51,660 --> 00:55:53,729 Well, there certainly are a lot of 1575 00:55:53,730 --> 00:55:55,979 issues with elections in the U.S. 1576 00:55:55,980 --> 00:55:57,329 and in many other countries. 1577 00:55:57,330 --> 00:55:59,219 And as you say, those are a number of 1578 00:55:59,220 --> 00:56:01,509 other ways that people's votes are 1579 00:56:01,510 --> 00:56:03,869 sometimes being suppressed 1580 00:56:03,870 --> 00:56:06,539 or are not being fairly counted. 1581 00:56:06,540 --> 00:56:09,089 Our expertize lies in the technology, 1582 00:56:09,090 --> 00:56:11,489 and so that's the primary area 1583 00:56:11,490 --> 00:56:14,549 of focus for me and for Matt. 1584 00:56:14,550 --> 00:56:16,619 But we often do get to work with 1585 00:56:16,620 --> 00:56:19,139 other election integrity advocates 1586 00:56:19,140 --> 00:56:21,459 who are concerned with, and we're 1587 00:56:21,460 --> 00:56:23,489 quite often in this wider spectrum of 1588 00:56:23,490 --> 00:56:24,449 issues. 1589 00:56:24,450 --> 00:56:26,459 But I agree with you, the politics and 1590 00:56:26,460 --> 00:56:28,769 technology here in 1591 00:56:28,770 --> 00:56:31,799 election security are so unfortunately 1592 00:56:31,800 --> 00:56:33,209 intertwined. 1593 00:56:33,210 --> 00:56:34,769 I don't think it should be a question for 1594 00:56:34,770 --> 00:56:37,409 anyone, whether our votes are 1595 00:56:37,410 --> 00:56:39,209 going to be checked to make sure they 1596 00:56:39,210 --> 00:56:40,769 weren't hacked. We should all want to 1597 00:56:40,770 --> 00:56:42,539 make sure they're checking their checked, 1598 00:56:42,540 --> 00:56:44,789 whatever our political affiliation. 1599 00:56:44,790 --> 00:56:46,769 But unfortunately, there's always 1600 00:56:46,770 --> 00:56:48,329 resistance from whoever the apparent 1601 00:56:48,330 --> 00:56:49,380 winner is. Thank you. 1602 00:56:50,910 --> 00:56:52,829 OK. Microphone six, please. 1603 00:56:54,600 --> 00:56:55,799 Thank you very much for your talk. 1604 00:56:55,800 --> 00:56:57,989 I've got one question that you didn't 1605 00:56:57,990 --> 00:57:00,569 seem to approach was the question 1606 00:57:00,570 --> 00:57:03,329 of voter suppression, 1607 00:57:03,330 --> 00:57:05,039 and in particular, hacking can mean an 1608 00:57:05,040 --> 00:57:05,969 awful lot of things. 1609 00:57:05,970 --> 00:57:07,709 There were an awful lot of machines in 1610 00:57:07,710 --> 00:57:09,959 Detroit that just didn't work or 1611 00:57:09,960 --> 00:57:11,610 were mis calibrated. 1612 00:57:13,470 --> 00:57:15,239 Could you approach that and how much? 1613 00:57:16,560 --> 00:57:18,959 How much can you contribute to fixing 1614 00:57:18,960 --> 00:57:21,479 that part of the system? 1615 00:57:21,480 --> 00:57:22,439 Well, we did sort of. 1616 00:57:22,440 --> 00:57:24,209 We touched on some of the systemic 1617 00:57:24,210 --> 00:57:25,800 failures, particularly in Detroit. 1618 00:57:27,030 --> 00:57:29,339 And, you know, I think that the solutions 1619 00:57:29,340 --> 00:57:30,749 to those kinds of failures and end of 1620 00:57:30,750 --> 00:57:32,609 things like voter suppression due to lack 1621 00:57:32,610 --> 00:57:34,769 of funding in a district are solved 1622 00:57:34,770 --> 00:57:35,939 by the same problem. 1623 00:57:35,940 --> 00:57:36,989 But you know, they have the same 1624 00:57:36,990 --> 00:57:39,299 solution. You provide funding 1625 00:57:39,300 --> 00:57:41,219 to buy better voting machines and so on 1626 00:57:41,220 --> 00:57:42,569 and so forth. 1627 00:57:42,570 --> 00:57:44,669 Beyond that, it's much more 1628 00:57:44,670 --> 00:57:46,469 complicated. And, you know, I'm not 1629 00:57:46,470 --> 00:57:47,470 really sure. 1630 00:57:48,600 --> 00:57:50,099 OK, all right. Microphone two, please. 1631 00:57:51,390 --> 00:57:53,009 Thank you for your for your work and 1632 00:57:53,010 --> 00:57:54,299 talk. 1633 00:57:54,300 --> 00:57:56,369 Moving away from the concrete 1634 00:57:56,370 --> 00:57:58,439 Donald Trump scenario, you said, 1635 00:57:58,440 --> 00:58:00,569 most to all, winners 1636 00:58:00,570 --> 00:58:02,729 oppose recounts. 1637 00:58:02,730 --> 00:58:05,759 So the question is if they can be 1638 00:58:05,760 --> 00:58:08,159 confident that they would be confirmed 1639 00:58:08,160 --> 00:58:09,209 as a winner? 1640 00:58:09,210 --> 00:58:11,400 Why do they oppose the recount? 1641 00:58:12,570 --> 00:58:14,759 Well, it's a good question, and this 1642 00:58:14,760 --> 00:58:16,829 has been my empirical observation in 1643 00:58:16,830 --> 00:58:18,269 every country where I've worked on 1644 00:58:18,270 --> 00:58:19,199 election security. 1645 00:58:19,200 --> 00:58:21,209 And I think the reason is because even 1646 00:58:21,210 --> 00:58:23,369 the Mirror suggestion that they might 1647 00:58:23,370 --> 00:58:25,439 have been elected unfairly, 1648 00:58:25,440 --> 00:58:27,599 the mere possibility that a recount 1649 00:58:27,600 --> 00:58:29,789 would expose any problems. 1650 00:58:29,790 --> 00:58:31,859 Well, it's worth the time and 1651 00:58:31,860 --> 00:58:33,749 effort of their lawyers opposing it to 1652 00:58:33,750 --> 00:58:37,049 try to make that possibility smaller. 1653 00:58:37,050 --> 00:58:39,359 I think people seek to move past 1654 00:58:39,360 --> 00:58:41,759 an election so quickly after 1655 00:58:41,760 --> 00:58:43,079 the votes are counted. 1656 00:58:43,080 --> 00:58:45,029 The first time that they're really there 1657 00:58:45,030 --> 00:58:47,129 isn't time to think about it, to check 1658 00:58:47,130 --> 00:58:49,229 the evidence to dot the i's 1659 00:58:49,230 --> 00:58:50,159 and cross the T's. 1660 00:58:50,160 --> 00:58:52,769 And that's that demand for instant. 1661 00:58:52,770 --> 00:58:54,869 Moving on is part of the problem. 1662 00:58:54,870 --> 00:58:56,879 It really makes the whole issue so much 1663 00:58:56,880 --> 00:58:58,169 more difficult to deal with 1664 00:58:58,170 --> 00:58:59,879 technologically. 1665 00:58:59,880 --> 00:59:00,839 Thank you. 1666 00:59:00,840 --> 00:59:03,119 OK, so we only have very little amount 1667 00:59:03,120 --> 00:59:05,129 of time left. So please be quick with 1668 00:59:05,130 --> 00:59:06,130 your questions, Mike. One. 1669 00:59:07,140 --> 00:59:09,119 Do you think there's value in end to end 1670 00:59:09,120 --> 00:59:12,089 verifiable elections and 1671 00:59:12,090 --> 00:59:13,379 cryptographic techniques to ensure 1672 00:59:13,380 --> 00:59:14,189 integrity? 1673 00:59:14,190 --> 00:59:16,139 Absolutely part of the solution, but the 1674 00:59:16,140 --> 00:59:18,149 complexity of the system already is so 1675 00:59:18,150 --> 00:59:19,259 difficult to deal with. 1676 00:59:19,260 --> 00:59:21,089 I'm not sure a local election officials 1677 00:59:21,090 --> 00:59:22,379 will be able to add any more. 1678 00:59:22,380 --> 00:59:24,149 We need to make it easier right now. 1679 00:59:24,150 --> 00:59:25,229 Audits are the best way to secure 1680 00:59:25,230 --> 00:59:26,230 election. 1681 00:59:27,210 --> 00:59:29,549 OK, unfortunately, we are out of time, 1682 00:59:29,550 --> 00:59:31,609 so please give another. 1683 00:59:31,610 --> 00:59:33,409 Huge applause for a meeting of it's busy.