NEW350+ rooms now have a live terminal, so you can learn by typing the commands, right in your browser.Learn in a live terminal. 350+ rooms.Try a lab♥ Codelivly is 100% free, no paywall. If it helped you, chip in to keep it running.100% free. Help keep it so.Support us
[>_] ROOM · Cybersecurity
Guided Lab: Identifying a Poisoned Model Artifact
A hands-on guided lab demonstrating how to detect hidden backdoors in AI models using activation clustering and adversarial probing.
Room tasks
1Lab Setup and Target Model Overview
2Scanning the Model File for Malicious Payloads
3Confirming and Documenting the Poisoning Vector
Frequently Asked Questions
What is a poisoned model artifact?
It is a trained model file that behaves normally on most inputs but contains a hidden backdoor triggered by a specific pattern. Attackers can distribute it through shared model repositories.
How does activation clustering find backdoors?
It examines the neuron activations of training samples and clusters them. Poisoned samples tend to form a separate cluster from clean ones, which reveals data that triggers the hidden behavior.
What does Neural Cleanse do?
Neural Cleanse searches for the smallest input perturbation that flips predictions to a target class for each label. An unusually small trigger for one class suggests a backdoor. You can practise this free in the lab.