Daniël

@alkatraz_200 @ferrebam @GrapheneOS I know, I was specifically addressing price. I hope gets better with more options, like the Motorola phones coming next year might get sold in different countries than Pixel.

If it is purely about security (I know other factors may be important to people, like the system being open source), iPhone is the next best thing. Yes, that's sad, but it is what it is.

@ferrebam @alkatraz_200 @GrapheneOS The last Android security bulletin almost 100 high/critical vulnerabilities in the main Android system components and over 80 in vendor-specific components (counted them approximately).

Waiting for weeks or months to fix them is terribly irresponsible (which unfortunately applies to most Android OEMs).

These are not theoretical issues, last week there was a chain of a WeChat + Android vulns that gave 0-click remote root access:

https://calif.io/research/weworm

@ferrebam @alkatraz_200 @GrapheneOS The companies that make 50-200 Euro phones are never going to make reliable monthly firmware, driver, and other security updates. It's nearly impossible at those margins.

Even companies with a huge sales volume, like Samsung, only offer quarterly updates on their budget models (A07, A17, A37, etc.).

Claiming that such phones are secure (even with GrapheneOS) would be a lie.

@ferrebam @alkatraz_200 @GrapheneOS Not everyone can afford a Pixel, but Pixel a-series are one of the cheapest phones on the market. They are often between 300-400 Euro new and can be had for 200-300 Euro second-hand or refurbished.

Anything below that price bracket will simply be a disaster security-wise, not only lacking MTE and a secure element, but also getting irregular to no security updates.

@engideer @GrapheneOS @ludonaut This is all besides the point that LLMs can find many vulnerabilities that fuzzers and traditional static analysis tools cannot (easily) find.

@engideer @GrapheneOS @ludonaut Inference will only get cheaper energy-wise as more and more specialized inference hardware is being made.

At any rate, there are many studies about this, e.g.: https://www.sciencedirect.com/science/article/pii/S2542435126001145

@engideer @GrapheneOS @ludonaut Frontier-scale inference uses 0.31 Wh/query. A 40 watt light bulb uses the energy of 130 LLM queries per hour. A query with large context is 13x more, but that’s still extremely marginal compared to a bunch of fuzzing machines running 24/7.

If you really want to stop destruction of the environment, stopping flying and becoming a vegetarian/vegan is much more effective than yelling at @GrapheneOS

@elle @GrapheneOS @dasmatus if you have watched reported vulnerabilities over the last few months, you'll see that there is a huge increase in the last few months due to LLMs. LLMs finds vulnerabilities that older methods such as static analysis and fuzzing didn't and don't find.

Since security is paramount to GrapheneOS and adversaries are using LLMs, the only way to defend is to use LLMs for auditing too.

LLMs are here to stay, let's boost open models to own the means of production.