Comrade Angles utilise witches.town. Vous pouvez læ suivre et interagir si vous possédez un compte quelque part dans le "fediverse".
Comrade Angles @Angle

Hmm. How long do y'all think it'll be before we start seeing neural networks used for hacking? :/

Moreover, how long till they get good enough at it that we start needing to use them for security? :/

@thamesynne Or even just dumb ransomware attacks that cripple hospitals and kill people? :/

@Angle we're already doing that. (I know of companies doing that.)

@joXn Huh. Can I get any examples? Or is it all secret? :/

@Angle let me see if I can get you some public references—I happen to be at an offsite that our chief scientist is also attending, and there's been a bunch of side discussion of machine learning on behavioral data for detecting intrusions. (I work for a cybersecurity company on our antimalware product.)

@Angle I thought this was about decryption but now I see you said hacking. idk.

@Shadow8t4 Decryption is a variety of hacking, IMO. I'd take that as an example. :/

@Angle well I'm not super into encrytion algorithms, but I would think even with the help of neural networks it wouldn't really be easy to make a system that would edge itself closer to a solution...

@Shadow8t4 Mmm, I would think so too. But is it really that much harder than making a system that plays go? :/

@Angle the thing with go is that it has predefined rules and a goal state. It's easy to tell a program what its legal moves are and what goal it should be trying to achieve. A neural network and learn this.

Decryption is more or less at this point somewhat of a brute forcing art. It's hard to just *know* what encryption algorithm was used to encrypt the data and harder to *know* when you're getting close to a decrption... At least, from what I understand... I could be completely wrong.

@Shadow8t4 Mmm, yeah fair enough. Still worries me though. :/