Hacker News Reader

Comment by Rd6n6 | Hacker News Reader

How would you use ai for cyber attacks in 5-10 years? Custom gpt3 optimized for phishing? Deepfakes for fake video calls? Deepfakes audio for fake phone calls? Captcha crushers by learning to fake fingerprints and fool trackers? Flood the internet with 10 million ai generated websites with fake content that cross references each other heavily to push fake news? In combination with identity theft as a step in a multi stage scheme? Mass swattings? Fabricating evidence of crimes?

What would you do to be safe from this?

View on HN

I can see politicians pushing all high quality faked videos as the real thing (without caring about origin) to bash their opponents. Fake evidence of crimes etc. Some societies are going to have a really hard time with this. The phishing angle should it come to pass would be solved with new systems of trust though it might be devastating at first. The fake generated websites feels like the current state of the internet ¯\_(ツ)_/¯

Authenticated videos maybe?

How would authenticating videos work practically?

The politician that was covertly recorded receiving bribes or ordering a murder surely will not say: "Yep, that's me on the video.".

You don't but you should already assume the internet is unsafe. The press is clickbait, wikipedia is a clique, and everything else is social media in disguise which implies political bias. The internet is a communication platform not a truth platform.

It isn’t enough to say that: Swatting and other fabricated crimes can still be targeted against someone who has no internet presence at all.

This is obviously technically true but so far every case I’ve ever heard about involved a dispute that began on the Internet.

Defense would probably involve a concept of 'strong identity'. A cheap way of saying it: 'Blockchain your life!'

At some point maybe people will finally be forced to stop believing everything they read on the internet

Reply on news.ycombinator.com