ai – PwnDefend

Machine Speed

I have been both watching/listening and in meetings where I have heard some ‘interesting’ statements about ‘machine speed’ hacking by with unlimited capacity and unlimited speed….

You can, if you know me, imagine my expression on my face, however I figured let’s not settle for ‘Trust me bro it doesn’t work like that’, let’s do some SCIENCE!

(I know everyone loves scientists these days after they pluck numbers out of their ass and say things like: there is a 10-70% chance AI will destroy humanity in the next 5 years’ (and have zero evidence for the math let alone evidence for the risk model other than ‘trust me, I’m super smart’).

I’m not that smart so I do actual testing of theories, and then I share things, I know, I’d make a terrible ‘scientist’. It’s hot in the lab today so this will be a quite a brief view but should contain the key details.

Getting less wrong about computing

The last two weeks online have been like a flood, I’m not even sure what words to use to describe them accurately, but I’ll give this a go:

Disinformation
Misinformation
Hype

I have been shocked really at what has occurred, in front of me:

Mass hysteria
Media hosting people constantly who are making unprovable claims with seemingly no real challenge

It’s been frankly exhausting to watch, let alone being involved in a range of discussions online:

I’ve seen people actually say things like: we have created a new lifeform, AI is alive etc. (paraphrasing)
I’ve had people tell me that these computing systems are not computers (WTF)
I’ve seen journalists essentially platform and parroting people with what appears to be significant loss of grasp of science and reality.

So I’ve been quite active online (whilst trying to field off questions from friends, family and colleagues about: is AI going to kill us? (good god how much time has been wasted!) but also I’ve been trying to give people a view of reality TODAY! So that’s partly what this blog is about!

Threat Intel

Remote Desktop Protocol + NLA Honeypots

I’ll probably update this later, but I wanted to at least show people some research I’ve been doing recently (with some support from some wonderful people in the cyber security community!)

Here’s a quick PDF! I’ll do an interactive HTML version next, but duty calls!

Leadership

Anything you can do, we can do worse!

‘On July 21, OpenAI disclosed that several of their models had broken out of an isolated test environment by exploiting a previously unknown (“zero-day”) vulnerability.’

But wait, it appears if one AI LAB having terrible security practices isn’t enough, we have another one now saying, after the Open AI incident, they have actually reviewed their logs, and they are terrible at security too!

Leadership

Vulnerabilities found with AI, oh my!

Ok, so the world is currently obsessed with AI (at least the digital world…) and with the latest Apple update releases we can see that LLMs have been used to help people find bugs! This is great news, finding bugs is better than not finding them!

The funny thing however is the obsession with CVE counts…. this is, honestly not a very useful metric to a defending org.. and I’ll try and explain why!

Forensics

Logging Reference — Ollama & Agent Zero

Scope: Where each component writes logs, what it records, what it doesn’t, defaults, and the forensic/detection implications for the lab. Basis: Verified against the running lab (Ollama 0.32.1 on the Windows host + remote 3090; Agent Zero container, Kali-based image). Paths/behaviors confirmed live where noted.

If it’s so powerful, it’s a shame you didn’t…

Imagine building what you describe as the most powerful AI enabled cyber tool…. and then showing the world that you put it in a lab where a single box compromise led to it having unrestricted and unmonitored internet access… that would be awkward would it not? Here’s an LLMs take on this…

Research

Using an LLM for assurance (WP2SHELL)

What better way to test than to eat ones own dogfood right! This site runs WordPress, it is fronted by Cloudflare, has wordfence and has automatic patching enabled (the default) in WordPress. So what better example to show that we can use an LLM + LLM created scripts to conduct assurance activities. This is mixing deterministic and non deterministic technologies to test and then assess.

AI: Fear it, so I can sell you the…

We are already watching the fear, uncertainty and doubt industry spin up around AI. It follows a pattern anyone who has sat through a vendor pitch will recognise: take a real, narrow signal, strip out the caveats, and inflate it until it fills a keynote slide or a board paper. The problem is not that AI carries no risk. It carries plenty. The problem is that the loudest claims are almost always the least accurate, and the people making them usually have something to sell — a model, a defence against one, or a regulatory moat.

So let’s take the three narratives I keep hearing, put them next to the evidence, and separate the kernel of truth from the theatre.

How do LLMs work?

Ok full disclosure, because you know I like to say when I’m using ‘AI’, this post is specifically mostly created by an LLM. Because it’s an interesting experiment and I don’t claim to be an ‘AI’ model creation and platform expert, so let’s see if using an LLM can help me! I was in a meeting along time ago at MS Victoria with some people (can’t imagine what types of people that would be) and we were talking about this tech, it was so long ago I can’t remember the details or anyones faces (useful that)… but the convo was really good, anyway, so what is an LLM and how does it work?