What will happen if all current encryption algorithms are cracked by AI?

General
1

Hi, I’m a non-professional and I have some questions about AI and encryption algorithms:
① What will happen if all the current encryption algorithms are cracked?
② Is it possible for AI to crack most or even all encryption algorithms in the short term (3~5 years)?
③ If there is such a possibility, and it is even more likely to happen, how should the entire society and decentralized platforms (such as IC) that rely on encryption algorithms respond?
④ Is the emergence of chatGPT good or bad for IC? And Why?
I just asked new bing to see if it can decompose the product of two prime numbers. I found out that it can’t do it yet, but I don’t rule out that it’s pretending not to be able to.
The plot of the last episode of Silicon Valley is that PiedPiper’s AI cracked the world’s strongest encryption algorithm, forcing the protagonist to kill it. Is it possible for this to happen in the short term?

2

Not a cryptographer… but my layman’s understanding is that “cracking an algorithm” takes two forms:

  1. The computational power of a machine is so large it can brute force its way to crack something. Crypto algorithms are designed to last for decades worth of computers getting better. Unclear (and unlikely IMO) this will change much due to AI. Maybe AI helps improve Moore’s law (but I’m skeptical on that tbh).

  2. There are side channel attacks in the “implementation” of a crypto algorithm. Theoretically, AI could help find these attacks faster than humans, but I have not heard of this has happening with any commonly used implementations of algorithms. It could certainly happen of course, and if AI gets sufficiently advanced maybe it can make looking for side channel attacks easier (inversely, it can also make fixing side channel attacks easier potentially). But this is pure speculation.

7 Likes
3

Hard to predict where AI is going, but I doubt very much it’s possible with the existing technology today. Not sure anyone could give you an answer for years or decades from now (with much confidence) given AI has a pattern of getting unexpectedly really good at surprising things but also being really bad at other surprising things.

7 Likes