Time to update the basics I think

Since AI has arrived on the scene and there are tons of articles telling all it can do, I would suggest you put AI into IPFire and have it supervise all things including monitoring the network for malware or any other issues. Of course set up the AI so that each deployment can be fine tuned to their specific use. Make all things adjustable. If you did this you would have the #1 firewall in the world.

1 Like

Your go! :grinning_face:

2 Likes

For this, use a special processor with 1(!) instruction only: DWIM. :slight_smile:

2 Likes

Haha, nice try.

A firewall is a deterministic engine that needs to make clear decisions that can be audited later. There will never ever be any AI near it. If so - and I am sure the marketing of some other vendors suggests that they are extensively using AI - you would lose all certifications and be non-compliant to pretty much every single Cyber Security certification that I know of.

This is all smoke and mirrors and we won’t engage in playing this game.

IPFire is rock solid the way it is and there is nothing from the AI camp that could improve the security of the firewall engine.

8 Likes

To stop a discussion about generating small scripts using AI before it begins, my opinion.

You can do this. And AI may give you suitable solutions. But verifying these programs may demand an effort comparable to doing the programming by hand. You need the same knowledge of programming for both tasks, verifying and programming.
The advantage of programming by a human is, the verifier can ask the programmer. You seldomly can ask an AI, why it generates this or other code.

We do not accept any AI-generated patches.

3 Likes

I know. But I wanted to clarify this aspect. You never know. :wink:

To be compatible with any of the licenses of the code and our contributor agreements, the author must own the copyright of the submitted contributing. If it is coming from a LLM, who knows who the copyright holder is?

There are a thousand technical reasons, and probably also another thousand legal reasons.

I agree with what everyone said here about AI and development. But a lot of the questions in the Forum concern configuration conundrums, which is a different thing.

And along with log reading and analysis, and CAKE QoS performance tuning guidance, a very small targeted AI could be helpful.

An own local IPFire AI could even be installed on a different machine, as long as there was the appropriate graphics card. Deployments are getting less expensive all the time.

Maybe even there could be an IPFire cloud AI that could answer questions from a growing body of knowledge - although not plugged into one’s own system.

Many of the questions in the Forum are very specific and of interest to a very small number of people. It wouldn’t be a good thing of course if the introduction of AI reduced IPFire community engagement - on the other hand it might end up that the Forum is focused on bigger policy and learning and use case issues that are of interest to more people.

Go IPFire! :tada:


I just realized that the AI topic had been raised in 2023 and provoked quite a discussion:

Interesting topic, am a little curiouse what has been meant in here :slight_smile: . Was ML (Machine Learning), DL (Deep Learning) or LLM (Large Language Models) the point of interest ?

Best,

Erik

P.S.:

There is meanwhile an updated version available to DWIS with 2(!) instructions :wink: .