HACKING THE MACHINES – COVERAGE FINALLY!
A recent spate of articles and audio programs is acknowledging that the manipulation of election results using electronic voting equipment is a distinct possibility. NPR, Wired, the Washington Post, Science Friday, and most comprehensively – an almost 9000 word article at Politico, are all reporting on the vulnerabilities of the machines that we vote on. The networks have joined the chorus with coverage at ABC, NBC and CBS. It’s a welcome sight to see this coverage, and it’s also about time. A small army of election integrity activists, computer scientists and journalists like Brad Friedman and Victoria Collier have made the point repeatedly over the last ten years; Collier most emphatically in an eye-opening 2012 article in Harper’s Magazine.
I have been one of those voices. In 2008, I released a documentary Holler Back – [not] Voting in an American Town. As part of the filming, I recorded the hack of the AccuVote-TS voting machine that is described in the Politico article. The hack is by then-graduate students Alex J.Halderman and Ari Feldman, working with Professor Ed Felten at Princeton. The documentary clip is 5 minutes long and well worth watching. It contains back and forth cuts between Halderman and a Lehigh County Pennsylvania election official, where Halderman explains how they overcame every single one of Lehigh County’s security measures.
Lehigh County Former Chief of Staff Levi Price, “We changed the passwords for each election.”
Halderman, “The password didn’t stop us for more than a few seconds.”
Lehigh County, “You have to have the technological capability of altering the software.”
Halderman,“The memory card is just a standard off-the-shelf memory card that you use in your digital camera. The smart cards you can buy over the internet.”
Lehigh County: “We have 700 separate machines. Each of those 700 separate machines is locked with a different key.”
Halderman, “You don’t need to break all 700 locks. Because a viral attack is possible, all you need to do is infect one or a small number of machines. And then the virus and your malicious vote-stealing software will spread from machine to machine during normal election procedures.
My narration: But what if we had a paper trail for the touch-screen machines. Would that make them more reliable?
Brad Friedman, (of Bradblog.com) “There is no reason to trust a touchscreen or DRE system with or without a paper trail. Period. You can hack the paper trails as easily as you can the internal numbers.”
That was in 2008.
HACKING IS NOT THEORETICAL
Which brings me to my next point: while the coverage is welcome, there are still some problems with the current narrative.
One problem with these stories is that they all take the position that this is a theoretical issue that they are concerned could happen. From the Politico article, “’You would be hard pressed to find an example of our voting systems ever being hacked in a real election environment, as opposed to that of a hack attempt inside of a laboratory environment in which zero real world physical election processes are utilized,’ writes Kathy Rogers, a spokesperson with ES&S, in an email, and correctly so—it’s never been proven that an election was deliberately hacked.”
This is actually not true. As early as 1994, the election of Nelson Mandela in South Africa was successfully attacked through the computers. According to Peter Harris, the head of the official election monitors, "The electronic count was compromised by a hacker who went in and multiplied the vote. The electronic count was then closed down. It stopped. That's when the results stopped going to South Africa and the rest of the world." The BBC reports that, “The secret manual system saved the day.”
If hacking the voting equipment is as easy as buying a memory card and a mini-bar key on the internet, and allows one control of local, state, and federal budgets it is not hard to imagine that it is already happening. The media may suffer from a lack of imagination, but hackers do not.
Our research does not prove that elections are being hacked. But the reported totals do not look accurate when analyzed statistically, and after extensively looking for other causes, we conclude that the manipulation of the count is the most likely explanation. The data indicates this is already occurring, and has in all likelihood been happening for years in both Democrat and Republican primaries, as well as general elections.
OPTICAL SCAN MACHINES ARE VULNERABLE
One concern about the current narrative is that some security experts are recommending optical scan machines as the solution to our election security woes. Our research indicates that computerized elections with a paper trail are as vulnerable as those without.
We performed a detailed statistical analysis of New York, a state that votes on paper, and then runs the ballots through optical scan machines. According to those familiar with the process, the results of the optical scan machines are all centrally tabulated by computer. In an interview Professor Halderman confirmed that central tabulation is a vulnerable security opening, where results could be assigned new percentages. “If you’re manipulating the central counting systems, then you can make an overall adjustment.” Asked whether it’s possible to get access to the central counting software, he replied, “Probably. It’s been our experience in the last decade with the Diebold systems, for instance, that it was certainly possible in the machines and counting systems we examined.”
Voting on paper is necessary, but if we want to be confident the results are accurate, it is also necessary to count the votes by hand. This is a conclusion that most of the European democracies have come to. As Ben Wofford points out in the Politico article, “Eight countries in Europe that once flirted with digital voting have seen six go back to paper; Britain counted its Brexit votes by hand.”
Below is the segment of our report showing how problematic the New York Democratic 2016 primary results are. For those who have not read the full report we found that a significant difference in the percentages that the candidates receive in small and large precincts is a cause for concern. There is no demographic explanation that is large enough to account for these variations. Furthermore, the variations are consistently helping or hurting certain candidates, ruling our random error as the problem.
New York State — Multiple Issues Raise Serious Concerns
In the 2016 New York Democratic presidential primary, the statistical patterns of certain counties are irregular (Figure 1). Both Richmond County (Staten Island) and Kings County (Brooklyn) reveal a strong correlation between precinct size and candidates’ percentages.
Fig. 1 — 2016 NY Dem pres primary. Richmond and Kings Co., show irregular statistical patterns
Columbia County, New York — a hand-counted county — offers a good point of comparison to the above graphs (Figure 2).
Fig. 2 — 2016 NY Dem pres primary
Columbia Co. - this hand-counted county has an expected statistical pattern
In this graph, there is a large degree of fluctuation both up and down on the left side of the graph. Then, by about 1,300 votes, the graph settles into a fairly even straight line, with small, random fluctuations. It maintains a basically flat line through all of the largest precincts. This is a very normal-looking CVT (cumulative vote tally) graph. Unlike the Richmond County, Kings County and Bronx County graphs, Clinton and Sanders have approximately the same percentages in the small and large precincts.
Earlier in the paper we demonstrated that hand-counted and machine-counted ballots in Kings County give different percentages for the candidates (Figure 3).
Fig. 3 — Hand-counted ballots show
a consistently higher return for Sanders in the 2016 New York presidential primary
Graph by Anselmo Sampietro
The suspect CVT graph in Kings County reinforces the findings of irregular election results presented earlier, and paints an increasingly troubled portrait of potentially compromised vote totals. But there is another odd factor in this state’s results.
Precise Percentages
We are showing the New York City graphs (Figure 1) to illustrate a very specific point. The data supports the idea that the overall state vote totals are being massaged to achieve a predetermined percentage.
The final reported totals in New York state were almost exactly 58% Clinton to 42% Sanders. It is necessary to go to the third decimal digit to see a difference: 57.995 versus 42.005.
Doug Johnson Hatlem reported on this issue, pointing out that, “The overall results in New York, as announced on election night, deviated from a perfect 58–42 split by 0.005345. That’s 97 votes out of over 1.8 million.”
In Kings County, the reported totals were almost exactly 60% Clinton, to 40% Sanders: (59.72% Clinton, 40.27% Sanders.) The .27 difference is caused by about 800 votes out of 300,000. In the Bronx, the percentages were almost precisely 70% to 30% (69.59% Clinton to 30.41% Sanders), the difference being just 616 out of 151,908 total votes. The Bronx County CVT graph is also irregular (Figure 4).
Fig. 4 — 2016 New York Democratic presidential primary
The Bronx County graph shows a strong correlation between precinct size and candidate percentage
New York City already had two unusual pieces of data giving cause for concern:
- The difference between hand-counted and machine-counted results
- Irregular statstical graphs in multiple counties
Add to that:
- Precise percentages in the total election results could be an indication that the results have been manipulated to achieve a specific percentage for the candidates.
These are three separate but reinforcing facts, illustrating why the totals for this state are suspect.
COUNTING BY HAND - THE BEST SOLUTION
Our recommendation is
- Paper ballots marked by the voter, counted by hand in the precinct with secure chain of custody and transparent processing that is open to scrutiny by the public and the media.
In this age of speed and automation many people cringe at the thought of sitting down at a publicly-monitored table and counting the votes one by one. But there are well-developed protocols for counting votes by hand publicly at each precinct, immediately following the close of polls, with good chain of command security, that make this a feasible enterprise. The process would be open to the media and could make for some very dramatic television – especially in a close race. People counting the votes by hand could give the networks hours and hours of dramatic air time to speculate about the possible outcomes.
Yes, it’s true, most of us use a calculator to add 17 + 24, but there are still a few things we would never want a machine to do. Would you want a machine to read a bedtime story to your child? Would you want a machine to make the final confirmation that a loved one was dead? Would you want a machine to determine whether or not we engage in nuclear warfare? After all, that is the ultimate decision we are deciding at the ballot box. Collectively we are determining who will make crucial decisions for us at the moment of crisis. It’s worth the extra time and effort to make sure we get it right.