The only way to Defeat Deepfakes is with Deepfakes

I’m not a Russian agent. I’m not planning to attack a US election (sounds like a glamorous job, but it’s not for me). But if I were, and I worked in the disinformation division, I’d have my GPUs spinning on the freshest Deepfakes money could buy.

I’d be impersonating the Democratic hopefuls, the squad, and who-knows-who else. The videos will be carefully crafted to damage these politicians, lose them votes and discredit the Democratic party in general.

We’ll have a DeepFake showing Elizabeth Warren calling American factory workers “a thing of the past”. We’ll have a snappy video of Ilhan Omar’s hot mic calling for Sharia Law and Joe Biden agreeing. We’ll have Marianne Williamson suggesting we sell Florida to the Moon People.

But, I’d not release a single video until we get close to November 3rd 2020. Why? Because the longer news sources have to discredit the videos the less impact the videos will have. Just like “photoshopped” became a universal term and everyone became skeptical of doctored photos, the same will happen with videos and audio. As George W. Bush said, “Fool me twice, can’t get fooled again”. It’ll just take time.

Therefore, the best way to blunt the impact of Deepfakes is to make TONs of them. NOW.

If we start releasing many election-interfering Deepfakes today, by the time November 3rd 2020 rolls around they will be old news. Everyone will be bored of them and assume these videos are fake.

If we let the attackers have their way, the only real malicious videos will be released too close to the election to be discredited.

We can’t prevent Deepfakes from spreading

Social media companies are not going to be able to stop these videos from spreading. I know the folks on the Twitter health team, and I know they are working on it and they are very smart, but getting a high-precision detection system in place on Facebook and Twitter and Instagram and Whatsapp is going to be a stretch before the election.

There are successful methods to detect Deepfakes, sort of — I saw this great talk by Mark Price and Matt Price at Blackhat 2019: Playing Offense and Defense with Deepfakes — and they describe effective detection algorithms. However, they also conclude that low-resolution videos are very difficult to distinguish, and regardless the detection methods are far from 100% accurate, and therefore will be hard to deploy.

→So Deepfakes will all be low resolution.

Deepfakes and other disinformation campaigns will be used to try to sway our next election. There’s no way to stop it. Let’s get on the offense now to blunt their impact.

Who’s in? Let’s start releasing tons of Deepfakes so they become old news.