Trojans in Artificial Intelligence (TrojAI) Final Report

TL;DR

This report summarizes the TrojAI program's efforts to understand, detect, and mitigate malicious backdoors in AI models, highlighting new detection methods, challenges, and lessons learned in AI security.

Contribution

It introduces foundational detection techniques for AI Trojans, evaluates their effectiveness, and discusses ongoing challenges in securing AI systems against backdoor threats.

Findings

Detection methods via weight analysis and trigger inversion

Performance metrics of Trojan detectors

Prevalence of natural Trojans in models

Abstract

The Intelligence Advanced Research Projects Activity (IARPA) launched the TrojAI program to confront an emerging vulnerability in modern artificial intelligence: the threat of AI Trojans. These AI trojans are malicious, hidden backdoors intentionally embedded within an AI model that can cause a system to fail in unexpected ways, or allow a malicious actor to hijack the AI model at will. This multi-year initiative helped to map out the complex nature of the threat, pioneered foundational detection methods, and identified unsolved challenges that require ongoing attention by the burgeoning AI security field. This report synthesizes the program's key findings, including methodologies for detection through weight analysis and trigger inversion, as well as approaches for mitigating Trojan risks in deployed models. Comprehensive test and evaluation results highlight detector performance, sensitivity, and the prevalence of "natural" Trojans. The report concludes with lessons learned and recommendations for advancing AI security research.