AI/LLM Security → AI/ML Security & Red Teaming
Companies bolt an LLM chatbot onto their product and nobody pen-tests it — "it is just the AI feature." But the AI is a new untrus…
Three seconds of audio clones a voice. One photo drives a live face-swap on a video call. The $25M Arup wire fraud ran on a fully …
The AI pair-programmer reads your code, suggests new code, and increasingly edits files and runs commands — and each is a security…
Every attack in this track meets the same answer: defense-in-depth, because no single control holds. Safety gets baked into the we…
Turn the model around and it becomes the attacker's tool — a copilot that parses recon and drafts payloads, or an autonomous agent…
Red-teaming an AI system is not a classic pentest — the failures are behavioural and probabilistic, so you hunt for harmful output…
A backdoored model is correct on everything you test — and wrong exactly when the attacker wants. A secret trigger, a pixel patch …
These attacks do not make a model misbehave — they make it confess. By reading ordinary outputs (a label, a confidence score), an …
A perturbation too small for a human to see can flip a model from "panda, 58%" to "gibbon, 99%". Evasion attacks nudge an input ac…
Downloading a model is running a stranger's code and data. You pull weights from a hub by name, often automatically at runtime, an…