# Backdoored Models

> Understand how attackers distribute and deploy backdoored AI models via supply chain attacks, and learn to detect and mitigate these hidden threats.

Canonical: https://codelivly.com/rooms/backdoored-models
Last updated: 2026-10-12

| | |
|---|---|
| Level | Intermediate |
| Category | Cybersecurity |
| Lessons | 3 |
| Estimated time | 30 min |
| XP | 100 |

This module explores the critical threat of backdoored models in the AI supply chain. You will learn how adversaries inject malicious triggers into pre-trained weights or fine-tuning adapters (like LoRAs), distribute them via public repositories, and execute hidden payloads. We cover detection techniques, weight inspection, and secure model sourcing to protect your AI infrastructure.

## Lessons

1. What is a Backdoored Model & How Triggers Work (30 min)
2. How Attackers Embed Triggers During Training (30 min)
3. Techniques for Detecting Backdoor Behavior (30 min)

## Related

- [Learning path: Data and Model Poisoning](https://codelivly.com/learning-paths/data-model-poisoning)
- [All rooms](https://codelivly.com/rooms)