January 16, 2017

My testimony before the House Subcommittee on IT

I was invited to testify yesterday before the U.S. House of Representatives Subcommittee on Information Technology, at a hearing entitled “Cybersecurity: Ensuring the Integrity of the Ballot Box.”  My written testimony is available here.  My 5-minute opening statement went as follows:

My name is Andrew Appel.  I am Professor of Computer Science at Princeton University.   In this testimony I do not represent my employer. I’m here to give my own professional opinions as a scientist, but also as an American citizen who cares deeply about protecting our democracy.

My research is in software verification, computer security, technology policy, and election machinery.  As I will explain, I strongly recommend that, at a minimum, the Congress seek to ensure the elimination of Direct-Recording Electronic voting machines (sometimes called “touchscreen” machines), immediately after this November’s election; and that it require that all elections be subject to sensible auditing after every election to ensure that systems are functioning properly and to prove to the American people that their votes are counted as cast.

There are cybersecurity issues in all parts of our election system:  before the election, voter-registration databases; during the election, voting machines; after the election, vote-tabulation / canvassing / precinct-aggregation computers.  In my opening statement I’ll focus on voting machines.  The other topics are addressed in a recent report I have co-authored entitled “Ten Things Election Officials Can Do to Help Secure and Inspire Confidence in This Fall’s Elections.”

In the U.S. we use two kinds of voting machines: optical scanners that count paper ballots, and “touchscreen” voting machines, also called “Direct-Recording Electronic.”  Each voting machine is a computer, running a computer program.  Whether that computer counts the votes accurately, or makes mistakes, or cheats by shifting votes from one candidate to another, depends on what software is installed in the computer.  We all use computers, and we’ve all had occasion to install new software.  Sometimes it’s an app we purchase and install on purpose, sometimes it’s a software upgrade sent by the company that made our operating system.  Installing new software in a voting machine is not really much different from installing new software in any other kind of computer.

Installing new software is how you hack a voting machine to cheat. In 2009, in the courtroom of the Superior Court of New Jersey,  I demonstrated how to hack a voting machine.  I wrote a vote-stealing computer program that shifts votes from one candidate to another.   Installing that vote-stealing program in a voting machine takes 7 minutes, per machine, with a screwdriver.  I did this in a secure facility and I’m confident my program has not leaked out to affect real elections, but really the software I built was not rocket science — any computer programmer could write the same code.  Once it’s installed, it could steal elections without detection for years to come.

Voting machines are often delivered to polling places several days before the election—to elementary schools, churches, firehouses.  In these locations anyone could gain access to a voting machine for 10 minutes.  Between elections the machines are routinely opened up for maintenance by county employees or private contractors.  Let’s assume they have the utmost integrity, but still, in the U.S. we try to run our elections so that we can trust the election results without relying on any one individual.

Other computer scientists have demonstrated similar hacks on many models of machine. This is not just one glitch in one manufacturer’s machine, it’s the very nature of computers.

So how can we trust our elections when it’s so easy to make the computers cheat?  Forty states already know the answer:  vote on optical-scan paper ballots.  (My written testimony clarifies this statement.)  The voter fills in the bubble next to the name of their preferred candidate, then takes this paper ballot to the scanner—right there in the precinct—and feeds it in.  That opscan voting machine has a computer in it, and we can’t 100% prevent the computer from being hacked, but that very paper ballot marked by the voter drops into a sealed ballot box under the opscan machine.  Those ballots can be recounted by hand, in a way we can trust.

Unfortunately, there are still about 10 states that primarily use paperless touchscreen voting computers.  There’s no paper ballot to recount.  After the voter touches the screen, we have to rely on the computer—that is, we have to rely on whatever program is installed in the computer that day—to print out the true totals when the polls close.

So what must we do?  In the near term, we must not connect the voting machines to the Internet.  The same goes for those computers used to prepare the electronic ballot definition files before each election, that are used to program the voting machines—that is, we must not connect the voting machines even indirectly to the Internet.  Many able and competent election administrators already follow this “best practice.”  I hope that all 9000 counties and states that run elections follow this practice, and other security best practices, but it’s hard to tell whether they all do.

These and other best practices can help protect against hacking of voting machines by people in other countries through the Internet.  But they can’t protect us from mistakes, software bugs, miscalibration, insider hacking, or against local criminals with access to the machines before or after elections.  So what we must do as soon as possible after November is to adopt nationwide what 40 states have already done: paper ballots, marked by the voter, countable by computer but recountable by hand.

In 2000 we all saw what a disastrously unreliable technology those punch-card ballots were.  So in 2002 the Congress outlawed punch-card ballots, and that was very appropriate.  I strongly recommend that the Congress seek to ensure the elimination of paperless “touchscreen” voting machines, immediately after this November’s election.

Which voting machines can be hacked through the Internet?

Over 9000 jurisdictions (counties and states) in the U.S. run elections with a variety of voting machines: optical scanners for paper ballots, and direct-recording “touchscreen” machines.  Which ones of them can be hacked to make them cheat, to transfer votes from one candidate to another?

The answer:  all of them.  An attacker with physical access to a voting machine can install fraudulent vote-miscounting software.  I’ve demonstrated this on one kind of machine, others have demonstrated it on other machines.  It’s a general principle about computers: they run whatever software is installed at the moment.

So let’s ask:

  1. Which voting machines can be hacked from anywhere in the world, through the Internet?  
  2. Which voting machines have other safeguards, so we can audit or recount the election to get the correct result even if the machine is hacked?

The answers, in summary:

  1. Older machines (Shouptronic, AVC Advantage, AccuVote OS, Optech-III Eagle) can be hacked by anyone with physical access; newer machines (almost anything else in use today) can be hacked by anyone with physical access, and are vulnerable to attacks from the Internet.
  2. Optical scan machines, even though they can be hacked, allow audits and recounts of the paper ballots marked by the voters.  This is a very important safeguard.  Paperless touchscreen machines have no such protection.  “DRE with VVPAT” machines, i.e. touchscreens that print on paper (that the voter can inspect under glass while casting the ballot) are “in between” regarding this safeguard.

The most widely used machine that fails #1 and #2 is the AccuVote TS, used throughout the state of Georgia, and in some counties in other states.

[Read more…]

Security against Election Hacking – Part 2: Cyberoffense is not the best cyberdefense!

State and county election officials across the country employ thousands of computers in election administration, most of them are connected (from time to time) to the internet (or exchange data cartridges with machines that are connected).  In my previous post I explained how we must audit elections independently of the computers, so we can trust the results even if the computers are hacked.

Still, if state and county election computers were hacked, it would be an enormous headache and it would certainly cast a shadow on the legitimacy of the election.  So, should the DHS designate election computers as “critical cyber infrastructure?”

This question betrays a fundamental misunderstanding of how computer security really works.  You as an individual buy your computers and operating systems from reputable vendors (Apple, Microsoft, IBM, Google/Samsung, HP, Dell, etc.).  Businesses and banks (and the Democratic National Committee, and the Republican National Committee) buy their computers and software from the same vendors.  Your security, and the security of all the businesses you deal with, is improved when these hardware and software vendors build products without security bugs in them.   Election administrators use computers that run Windows (or MacOS, or Linux) bought from the same vendors.

Parts of the U.S. government, particularly inside the NSA, have “cyberdefense” teams that analyze widely used software for security vulnerabilities.  The best thing they could do to enhance our security is notify the vendors immediately about vulnerabilities, so the vendors can fix the bugs (and learn their lessons).   Unfortunately, the NSA also has “cyberoffense” teams that like to save up these vulnerabilities, keep them secret, and use them as weak points to break into their adversaries’ computers.  They think they’re so smart that the Russkies, or the Chinese, will never be able to figure out the same vulnerabilities and use them to break into the computers of American businesses, individuals, the DNC or RNC, or American election administrators.  There’s even an acronym for this fallacy: NOBUS.  “NObody But US” will be able to figure out this attack.

Vulnerability lists accumulated by the NSA and DHS probably don’t include a lot of vote-counting software: those lists (probably) focus on widely used operating systems, office and word-processing, network routers, phone apps, and so on.  But vote-counting software typically runs on widely used operating systems, uses PDF-handling software for ballot printing, network routers for vote aggregation.  Improvements in these components would improve election security.

So, the “cyberdefense” experts in the U.S. Government could improve everyone’s security, including election administrators, by promptly warning Microsoft, Apple, IBM, and so on about security bugs.  But their hands are often tied by the “cyberoffense” hackers who want to keep the bugs secret—and unfixed.  For years, independent cybersecurity experts have advocated that the NSA’s cyberdefense and cyberoffense teams be split up into two separate organizations, so that the offense hackers can’t deliberately keep us all insecure.   Unfortunately, in February 2016 the NSA did just the opposite: it merged its offense and defense teams together.

Some in the government talk as if “national cyberdefense” is some kind of “national guard” that they can send in to protect a selected set of computers.  But it doesn’t work that way.  Our computers are secure because of the software we purchase and install; we can choose vendors such as Apple, IBM, Microsoft, HP, or others based on their track record or based on their use of open-source software that we can inspect.  The DHS’s cybersecurity squad is not really in that process, except as they help the vendors improve the security of their products.  (See also:  “The vulnerabilities equities process.”)

Yes, it’s certainly helpful that the Secretary of Homeland Security has offered “assistance in helping state officials manage risks to voting systems in each state’s jurisdiction.”  But it’s too close to the election to be fiddling with the election software—election officials (understandably) don’t want to break anything.

But really we should ask: Should the FBI and the NSA be hacking us or defending us?  To defend us, they must stop hoarding secret vulnerabilities, and instead get those bugs fixed by the vendors.