Do you think AI will be able to crack GOS?

notnn

Title sums it up pretty much. Soon we'll have AI much smarter than humans (which in and of itself is worrisome enough). I'm hypothesizing a scenario where AI tries to gain access of our electronics. Do you think GOS will be able to withstand such an attack? Keep in mind I'm just your average user so I don't know if the question I make even makes sense.

Blastoidea

What, precisely, does “crack” GrapheneOS mean?

It’s your term, so I’d like to hear what it means to you.

notnn

Blastoidea You are right, as i said I'm just a casual user so I'm not really able to define it properly. A vague definition of what comes to mind is anything that is different than the state that the system of me and the device is now. I'm the only one in control of the device and only one having access to what is on the device, where it goes, what happens to it, if that makes any sense.

Blastoidea

Are envisioning someone taking it away from you, or confiscating it?

I may still be confused.

notnn

Blastoidea I was thinking of remote attacks but after spending some more time thinking I believe anything would be possible if it is more intelligent than us and out of our control. I don't know, maybe we are doomed and my phone should be the least of my worries.

6gsxdr3U

If you are talking about encryption at rest: AI won't magically break that.

notnn

6gsxdr3U Yeah but doesn't that require that we still understand what the AI is doing? What about 10 years from now, when it is much more advanced? What if it figures out stuff we could never even imagine. Food for thought I guess.

6gsxdr3U

notnn AI is not 'more intelligent' then us, and won't be so for the foreseeable future. AI is good for selective tasks, that humans struggle with, e.g. facial recognition, or querying data from massive knowledgebases (The LLaMa stuff that is very popular at the moment).
GrapheneOS FDE is implemented with AES256 (See FAQ for details), which is computationally secure. There won't be a magic algorithm that an AI comes up with that cracks AES. If there is a security flaw in FDE, then it would be in the implementation used (e.g. allowing for some side-channel attack). Currently there are there are none we are aware of.

Where Ai will and is more problematic is in terms of tracking peoples activities on websites and such. But that is not GrapheneOS specific.

Blastoidea

notnn
You worry about some strange stuff.

I really can’t seem to come up with a helpful reply to your question.

de0u

Blastoidea You worry about some strange stuff.

It was one considered weird for people to worry about surgeons having clean hands, now very much the other way.

Looking into random/novel ideas, including worries, can be productive.

Graphite

I guess AI is in the news and confusing enough for most people, that AI might be the misnomer people are using to describe other things.

Quantum Computing cracking encryption is the thing you're probably thinking of. And it's also paradoxically overrated and underrated.

6gsxdr3U

Graphite The problem about PQC is that its often incorrectly portrayed in the media. Quantum computers do pose a risk, but until they get more QBits, and especially less "noisy" QBits we won't see classic cryptographic schemes being broken for a couple more years. Also, the standardization process of PQ-resilient algorithms is already ongoing (See: https://www.keyfactor.com/blog/nist-announces-round-3-finalists-for-post-quantum-cryptography-competition/?hsa_acc=9535308306&hsa_cam=&hsa_grp=&hsa_ad=3767890226168001212&hsa_src=&hsa_tgt=&hsa_kw=&hsa_mt=b&hsa_net=adwords&hsa_ver=3&utm_content=dsa)

Eventually, in the future software and hardware will be migrated to these algorithms where suitable. But that takes time.

Lodonnell1392

I don't think it comes from AI, more quantum computing, maybe mixed with AI.

Think of it this way, AI is still running on traditional computers, okay lots at once, but still even millions of modern computers together couldn't break AES256 encryption.

I think your title is worded incorrectly, instead maybe can AI affect GOS security? And the answer I think is, it could? But GOS it's self is hardened above normal android, and android has the open source + the sheer power of Google backing it, plus the grapheneos team who also put alot of effort into the system.

So, I'm theory yes it could have some effect on GOS security, but I don't think it's as big of an issue currently, it would have to be a targeted attack also, the bigger threat is quantum computing, not ChatGPT however.

Hope this makes some sense :)

Phead

Well, I don't think that AI itself (from a mere computational point of view) is a risk, at least not as of now. BUT since AI is really good at collecting data it's absolutely possible that it might become useful in cracking weak passwords by means of social engineering (compiling data available on social media, news, job, what not). I don't know how I would train it to be good at this job but I think that's just my lack of criminal energy (also maybe intellect).