Voting security: policies versus technologies

Originally Published on NevadaNewsmakers.com, 8/15/2007 2:59:41 PM

The Sequoia voting system is among “the most tested, secure, accurate, auditable and accessible voting systems in our nation’s history,” according to a statement on the company’s website.  Either that, or it’s a  product that could have been better engineered by any U. C. Berkeley student who did well in an undergraduate computer security class taught by Professor David Wagner.  Wagner was a team leader for the recent “top-to-bottom review” of California’s voting system security, and he’s fairly confident that a serious student in his course would have avoided many of the design flaws that showed up in the Sequoia system review.

Sequoia seems equally confident in its product, even after a long list of successful hacks were performed on it during the review, which was ordered by the California Secretary of State.  Scrutiny by teams of computer scientists revealed numerous deficiencies in the system, and a level of security one would expect “in a system where security just doesn’t matter,” by David Wagner’s account.

Sequoia Voting Systems, maker of Nevada’s election equipment, called the tests “irresponsible and misleading.”   The company’s primary assertion is that they were performed under unrealistic conditions. Sequoia claims the results of the tests are distorted, because of failure to include a defense team — “people, procedures, and processes”  to protect the system while the so-called Red Team was making hay with its technical vulnerabilities. Security threats to the system have therefore been overstated.

This is the central question.  To what extent should election security be assured by the design of the equipment, and to what extent by the people who run the elections?  As a general practice, information systems are protected through layers of technical features combined with elements of personnel and policy management.  Both are necessary, and this is Sequoia’s point.

But over-reliance on people and procedures could be fatal to security for a poorly-engineered system, according to U.C. Berkeley’s Wagner.  The human element is crucial, but it’s not sufficient.

“One reason our results are so worrying,” says Wagner, “is that we found that every single one of the technological defenses could be breached.  Because the security of the software is so weak, you’re completely reliant on poll workers and election officials… They have to work much harder to adequately protect elections than they otherwise would.

“Our current (election) procedures were designed under the assumption that the software was a lot more secure than it actually is,” Wagner continues. “That’s troubling because any procedural mistake at any point can have serious consequences for the integrity of the election.”

On this point Wagner gets a certain oblique agreement from Sequoia. In hearing testimony, the company offered a point-by-point refutation of some of the vulnerabilities ascribed to it, but not always on technical grounds.  Often, Sequoia’s recommended “fix” for a technical attack on the system is a people fix, suggesting that the system would be invulnerable, so long as the election workers are well-trained, ever-vigilant, and not corruptible.

The dreaded “yellow-button attack,” for instance, which has been a major source of buzz among election watchdogs.  The Sequoia voting machine can be switched to manual mode by simply reaching around to the back of the machine and pressing the activation button. This would allow a voter to vote repeatedly, rather than just once.

Sequoia recommends turning the machines around, so the back of the units faces the poll workers. This assumes that the poll workers will be consistently attentive throughout a long election day, never busy with other duties, and most important, not distracted by an accomplice whose job it is to chat up the poll workers while the yellow-button attacker does his work.

Repeatedly, Sequoia invokes personnel, procedure, and policy as a defense for the system’s known vulnerabilities. Audits, training, anti-virus and anti-spyware updates, secure storage of equipment, camera surveillance. Such measures are important, says Wagner.  But they don’t mitigate the system’s blatant flaws.

“I’m a little concerned that the response seems to be deny, deny deny,” he says. The design problems need to be acknowledged, and then they can be addressed.

Wagner suggests that the focus on people and policy is a way for Sequoia to shift blame to county election officers. Elections managers in both states, though, seem to be in verbal lockstep with the vendor, however, taking their talking points directly from Sequoia’s own statements.

Computer security professor: Sequoia voting system would get a D or an F

Originally Published on NevadaNewsmakers.com, 8/13/2007 2:59:58 PM

“It would be like giving someone the key to the bank, and the combination to the safe, then leaving them alone for months. You shouldn’t be surprised if the money is gone when you come back.”  This is the analogy repeated separately by various people who discount California’s security review of its electronic voting systems.  By this, the critics mean that teams of computer scientists who were able to hack the systems were armed with source codes, and had “unfettered access to the machines” — this is the other oft-repeated term  – for the four-month duration of the experiment.

Critics also point out that the tests were conducted in a laboratory setting, which did not recreate actual election day conditions. On election day, they say, there would be another team playing defense, which might prevent hackers from compromising the election.

Skeptics include Nevada’s election officials.  The Silver State uses the Sequoia system, not that it would matter.  All the systems in California ‘s test were hacked, including Sequoia.  The “keys to the bank and combination to the safe” analogy has been used by Washoe County Voter Registrar Dan Burk, and the Secretary of State’s Elections Deputy Matt Griffin.

Wrong analogy, says Dr. Matt Bishop of U. C. Davis, who led the so-called Red Team that  produced an extensive list of ways to penetrate the Sequoia electronic voting system.  A closer parallel would be the crash-testing performed on cars by the National Highway Traffic Safety Administration.  NHTSA, he says, is evaluating the limits of the machinery, not the conditions under which cars move around on the road with real people behind the wheel.  It’s an extremely rare driver who pushes the pedal to the metal, and drives, head-on, into a brick wall.  But such “laboratory” tests are useful because they document the machine’s behavior when it’s subjected to a crash, and the physical damage to the occupants.

And yes, says Bishop, the Red Team did have Sequoia’s source codes, allowing it, in effect, a guided tour of the software design, and an insider’s knowledge of how the system performs its functions.   Bishop believes an organized conspiracy to commit electronic voting fraud would also have the source codes.

The codes most likely would be leaked or stolen, since most information security breaches are committed with the help of insiders.  But source codes might be obtained in a variety of ways. Voting machines have been sold on eBay, for instance, by government surplus agencies, notably to a curious geek whose purchase was chronicled in Wired Magazine.  He promptly took the machines apart to figure out how they function.

Most information is out there, Bishop says, if you dig deeply enough. He cites several incidents involving national security, where classified information was unintentionally posted in a place where it could be viewed online by the public. The information was removed, but the bell can’t be un-rung after it’s had thousands of page views.

Source codes were given to the University of California teams primarily to help them meet the deadline imposed by the California Secretary of State.  Four months is a short period for the kind of work requested, and the source codes made the work go faster.  Without the codes, Bishop says, the team could have discovered and performed the same list of hacks, it would simply have taken longer.

The California test resulted in decertification of the Sequoia system as it is currently used in that state. In the same document, there is a long list of conditions Sequoia must meet in order for its system to be used in California’s upcoming primary, and a long list of new procedures that must be developed by each election jurisdiction.  In short, the recertification conditions create a large and expensive headache for everyone involved, with a short window for completion.

At this point, the only certain use of the Sequoia in the California primary will be for the disabled. Access for the disabled was among the original objectives of the “Help America Vote Act,” the law that spawned universal implementation of electronic touchscreen machines. The machines are wheelchair accessible, and are equipped with a listen-only ballot and headphone jack for blind voters.  Secretary of State Debra Bowen has ordered a single machine in each precinct for HAVA disabled compliance.

Nevada officials say they’ve watched the California tests with interest, as has the entire national community of election managers.  Both Washoe’s Dan Burk and State Elections Deputy Matt Griffin say they’ve read the reports issued to California’s secretary of state.  Griffin says the information will be used at the state level by Nevada election committees who have been assigned to make recommendations for various aspects of the election process. One of the committees is an IT committee, which will presumably study the technical exploits from California’s tests. The findings of the tests will be used by another committee to develop enhanced training for elections personnel.

In general, though, Nevada’s focus seems to be on policy and procedure, not on the shortcomings of the Sequoia system itself, which Matt Bishop says he would grade D or F if it were  submitted as a project in one of his computer security classes at U.C. Davis.

Hack the Vote!

Originally Published on NevadaNewsmakers.com, 8/7/2007 1:01:24 PM

Hack the Vote! Part 1

The report might have been titled “A Thousand and One Ways to Compromise an Election.” It was prepared for the California Secretary of State by a team of computer scientists acting as white-hat hackers, conducting penetration tests on various brands of electronic voting machines used in California. It was a collection of reports, actually, and the one of interest to Nevadans could have been called “A Hundred and One Ways to Compromise the Sequoia Voting System.” That’s the system used in Nevada elections.

Although the members of the Source Code Team from U.C. Berkeley claim they didn’t have enough time for a thorough review, they nonetheless compiled an enormous list of ways in which the Sequoia system is deficient on every level from the back end system to the hardware in the precincts.

A few of the findings: The “smart cards” voters insert into the machines can be altered using a laptop, as can the “results cartridges” that hold the tally of all the day’s voting for each machine. It’s possible for results affecting one precinct or multiple precincts to be changed without detection.

The tamper-evident seals that alert poll workers to an attack on the hardware, or the machine’s internal firmware, are irrelevant, since “the  testers were able to gain access to the internals of the systems by, for example, unscrewing screws…. Similarly, plastic covers that were protected by seals could be pried open enough to insert tools that could manipulate the protected buttons without damaging the seals or leaving any evidence that the security of the system had been compromised.”

Features that limit control of the WinEDS database are easily circumvented, according to the report. This means the people at every level who are involved in back-end operation, such as preparing ballots, configuring voting machines, and counting the ballots, can assign roles to themselves and gain access at a level they were never intended to have.

The results were no surprise to the broader community of tech intelligentsia, which collectively uttered one word when the story broke. “Duh.”

The number and severity of vulnerabilities in the systems led Secretary of State Debra Bowen to decertify the Sequoia machines on Friday, to meet a deadline related to the California primary, and re-certify them under more stringent conditions.

Sequoia hosted a user conference last week, the timing of which, deliberately or not, coincided with the release of the California test results. Election officials from across the country, including our own, attended. Security experts say all the companies whose machines were tested have an interest in discrediting the results, and indeed, when the testers gave their report at a hearing in Sacramento, it was vendors and county election officials who squared off with them.

Some at the county level, where jobs are made more difficult by the findings, accuse Secretary of State Debra Bowen of staging the test as an exercise in political theatrics. They may be right, but that doesn’t minimize the significance of the tests.

——————————

 

Hack the Vote! Part 2

First, we have to lose this romantic notion about the nature of cybercrime, deeply embedded after only a decade or so of Hollywood portrayals.   The Lone Gunmen of “X-Files” fame, or Chloe and Morris of “24” tear into the keyboard as the clock ticks. They’re under intense pressure, with lives in the balance. Their brows furrow and beads of sweat appear on their upper lips. Within minutes — hours if it’s a tougher assignment — there’s a blinking message window on the screen, signaling success.

Then there’s the swaggering, teen-aged “because-I-can” hacker, who operates from a bedroom in his parents’ home, altering school records or dropping dirty pictures onto the websites of prominent people, for laughs. In another era, he’d have been vandalizing parking meters. He’s real, but he’s not committing election fraud. Yet.

More recently, the public has begun to grasp the common criminality of hacking, as eastern European syndicates using stolen credit card numbers have been revealed to be a major source of cybercrime. This crime, motivated by money, is the closer to the worry at hand, but still not quite right.

Here’s a chilling paragraph from the U.C. Berkeley report:

“The high stakes of many elections can provide rich incentives for illegal abuse. The resources of this study were quite small compared with those that might be available to an organized criminal conspiracy to commit election fraud. In our professional opinion, it is likely that a sophisticated adversary who sought to subvert an election by technical means would be able to duplicate, and perhaps exceed, the attacks reported here, even without open access to source code. Our study should be regarded, at best, as identifying obvious weaknesses that an attacker might attempt to exploit.”

Get the picture? Cyberattacks on elections would be painstakingly planned, with a possible interval of four years to get it right. And lots of money would be available to people with the requisite combination of know-how, financial need and political disgruntlement. With regard to the categories of cybercriminal for which we have a frame of reference, they would fall into the column marked “none of the above.”

The Reasonable Reporter, having produced a five-part series on electronic voting machine security in the summer of 2004, is in contact with some folks who have spent a lot of years studying voting machines. More to follow.