Alex Halderman knows the vulnerabilities in election technology and has advice for voters to minimize their personal risk. But why aren't we better securing our elections?
Key takeaways
- Election systems are not supposed to be connected to the internet, but testing has revealed that some are.
- The slowness of patching election systems is on a collision course with AI's ability to rapidly discover vulnerabilities.
- Hand-marked paper ballots remain the most secure election technology available.
President Trump has alleged many times that the 2020 election was "rigged." Democrats have often repeated the statement by the Election Infrastructure Government Coordinating Council Executive Committee that it was "the most secure in U.S. history."
The University of Michigan's Alex Halderman, who has spent two decades exploring the vulnerabilities of American voting systems, is tearing his hair out. While he affirms that there is no evidence that these flaws have been exploited, neither Democratic nor Republican administrations have used their law-making power to secure them, he says.
Halderman, the Bredt Family Professor of Engineering and a professor of computer science and engineering, discusses the most serious technological security issues in American elections. He also has advice for voters about the most reliable ways to vote.
President Trump's election security speech in July was widely criticized, but you argue it included important facts everyone should be paying attention to. What are they?
More interesting to me than the president's words were the documents that he declassified. One, in particular, is an assessment of the current state of election security by CISA, the Cybersecurity and Infrastructure Security Agency-the arm of the Department of Homeland Security responsible for election security. This assessment goes into a lot more detail than their previous communications about some of the problems observed in coordination with the states.
One of the central themes of reassurance that we've been giving to voters is that we don't have to worry about hacking because election systems are not connected to the internet. But between 2018 and 2024, CISA ran exercises with the states, assessing their security and even trying to hack into their network environments to see whether an adversary would be likely to do it.
In a lot of cases, CISA was able to hack into things that weren't supposed to be connected to the internet, but people forgot to completely disconnect them. Or the network was designed in a way that looked like it was disconnected, but it really wasn't. This underscores that there's more we need to do in terms of policy, patching and getting resources to the states to make sure that there isn't a path from the internet that a foreign adversary could take to disrupt our elections.
You've raised concerns about how AI changes the speed at which cybersecurity threats emerge. What does that mean for election technology?
In election security, one of the chronic problems is that it takes forever to patch vulnerabilities. Any change to the system needs to go through different kinds of approval, and updates have to be installed manually by people traveling out to the field. It's slow; it's expensive. For example, the state of Georgia is going to vote this fall with vulnerabilities that I discovered myself-including critical ones, some back in 2020-that have not yet been patched.
All of that is on a collision course with artificial intelligence. Over the last year, AI models have demonstrated a superhuman ability to discover and exploit vulnerabilities in all kinds of systems.
To keep up with AI, we need policy reform that ensures election technology vendors uncover security problems more effectively and states apply available security patches promptly. If you're going to run an election system that has tens of thousands of computers out in the field-which is how many states run their polling places-it takes a lot of resources and planning to make sure those systems are fully patched before every election event.
You have a simple prescription for secure elections. What is it?
The consensus of the scientific community is that voters should mark ballots by hand when they can. We should count those ballots initially by computer, and then we should audit every significant contest by having people go back and look at enough paper ballots to know that the outcome was right.
If you do those things, then there isn't room for a cyberattack, for hacking, or even for bugs to change the outcome of the election in an undetectable way-because we have verification being done by people in the physical world.
Does the Safeguard American Voter Eligibility Act do any of that?
The SAVE Act has very little to do with the security of election infrastructure. It is entirely about a different question: How do you determine whether voters are eligible to vote?
While there's very little evidence of noncitizens voting at any scale, and it's punished quite severely, it's difficult to prove that it doesn't happen more often. This bill is a response to this concern, which is held by many voters.
Voter eligibility is a classic example of a hard security trade-off in elections. We want every eligible voter to be able to vote, and we want no ineligible voters to vote. Things that make it harder for ineligible voters to cast votes-stricter ID requirements and the like-also make it more difficult for some eligible voters. Where you set that balance is ultimately a policy question rather than a security question, and different states have arrived at different answers.
The SAVE Act would uniformly require greater checks on eligibility across the country. However, it would leave the security of election technology in its current patchwork of strength and weakness.
As someone who knows the faults of our electoral systems better than most, do you still vote? How do you make sure your vote counts?
I love voting. Even if you're concerned about weaknesses in election systems, your vote is certain not to count if you don't vote at all.
If you can, fill out a paper ballot by hand and vote in person because that's the strongest way to make sure that it gets counted correctly. When you make a physical record, there's no potentially vulnerable computer sitting between you and the record of your vote. Voting in person also provides an opportunity for error correction. Say you picked too many choices in some race, which would cause those selections to be discarded-the ballot scanner can tell you that immediately.
If your state makes you vote on a touchscreen, take time to carefully review the printout, item by item. And reach out to your representatives-tell them that hand-marked paper ballots are the most secure election technology available, and you want your state using them.
Most of all, I think it's important for voters to educate themselves about the science behind election security so that they can be appropriately skeptical of claims from all sides. The existence of vulnerabilities isn't proof that an election was hacked, and the absence of overt signs of hacking doesn't prove an election system is secure.
