人工智能中的木马（TrojAI）最终报告
Trojans in Artificial Intelligence (TrojAI) Final Report
AI总结：
TrojAI计划旨在研究和应对AI木马威胁，通过权重分析和触发器反向分析等方法提升AI模型的安全性，并提出检测与缓解策略。
摘要：
美国情报高级研究项目计划局（IARPA）启动了TrojAI计划，以应对现代人工智能中出现的新漏洞：AI木马的威胁。这些AI木马是恶意的、隐藏的后门，故意嵌入到AI模型中，可能导致系统以意外方式失败，或允许恶意行为者随意劫持AI模型。这项多年度计划帮助厘清了该威胁的复杂性质，开创了基础检测方法，并识别出需要持续关注的未解决问题，这些问题是人工智能安全领域亟需解决的挑战。本报告综合了该计划的关键发现，包括通过权重分析和触发器反向分析进行检测的方法，以及在部署模型中减轻木马风险的途径。全面的测试和评估结果突显了检测器的性能、灵敏度以及“自然”木马的普遍性。报告最后总结了经验教训，并为推进人工智能安全研究提出了建议。
英文摘要：
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.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
