The sinister timing of deepfakes and the 2020 election

Education and legislation are needed to combat the significant threat of deepfakes.

The sinister timing of deepfakes and the 2020 election

TechRepublic's Karen Roby talked with Matt Price of ZeroFox about the type of deep fakes circulating and their potential impact in the election cycle. The following is an edited transcript of their interview.

Matt Price: Audio deepfakes are just synthetic audio usually trained on a certain person's voice, such as the CEO of a company, and then you're able to then feed text into that algorithm, which then generates the appropriate words using that person's synthetic voice.

Video deepfakes, on the other hand, those are videos where you usually copy someone's face onto either an actor or just replace that original person's face, making them say something they never really said. So this can be very short spans of time. The video could be maybe just five seconds of modified video, or it could be the entire video is modified. We've seen deepfakes that are several minutes in length.

SEE: How to secure your zoom conference line from hackers (free PDF) (TechRepublic)

Karen Roby: Expand for us on deepfakes during the election season and how concerned should we be?

Matt Price: We've seen proofs of concepts of deepfakes being released that could be used to influence the electorate. We haven't seen any deepfakes released in the wild that we think are genuinely malicious, not saying that they're deepfakes and trying to mask what they are. But really at the end of the day they could have huge influence. If a deepfake is dropped at the right time, say maybe two or three days before an election occurs, imagine the impact that could have if it goes viral? And within that two or three day period, it may not be enough time to go and conclusively prove that this video was a deepfake. That could sway the electorate then, enough to vote for a particular political candidate. That's really the huge question about deepfakes and how they could be deployed and how we would stop them before they went viral.

Beyond just deepfakes though, we've already seen the influence operations that happen today with texts and images. So deepfakes are really just another tool in that malicious actor's arsenal to conduct these influence operations.

Karen Roby: Are there different types of scams involving deepfakes in elections in particular?

Matt Price: In terms of elections at least, deepfakes, as I mentioned before, are just a small piece of the puzzle. There is a lot of stuff that already goes on today around elections that I don't think a lot of people are aware of. One of the big ones are imposters. There's a lot of people that go in and impersonate a political candidate, say we've caught a few with Bernie Sanders for example, and a couple with a President Trump. So those imposters go around, try to spread a slightly different message or they try to collect donations from people and obviously then disappear with the donations.

Similar to that, there's also scams where they're not trying to necessarily impersonate the political candidate, but they are trying to get money or personal identifiable information so they could perhaps later then try to break into a person's accounts.

Karen Roby: So the big question is, what do we do?

Matt Price: There's a number of different ways that this problem can be tackled, and I don't think any one by itself is a solution. One is just proactive policing on large websites, and particularly social media. Clearly designating that something is synthetic media or if the facts in the article cannot be proven, I think would go a long way toward just educating people about what they're reading.

The other piece is education. You have to be very aware and cognizant that what you're reading online may not be the truth. There is not necessarily a lot of policing with the content that gets put online and you never really know where it's coming from. Even if it supposedly states it's from, say, a large news organization, that could be obviously just synthetic and made up.

So those two things I think are critical. The other thing is legislation. At the end of the day, there are no laws that regulate what's going on with social media and websites today. They aren't responsible for the content that users put up there so they don't really have an incentive to go round and try to take down this synthetic content or even note that it is synthetic content. Coming up with some kind of political solution is the other piece of the puzzle as well.


Karen Roby talked with Matt Price of ZeroFox about deepfakes and their potential effect on elections. 

Also see