审计守门人：模糊测试“AI法官”以绕过安全控制

The research reveals a critical vulnerability in AI judges, which are large language models (LLMs) used as automated security gatekeepers. Researchers developed AdvJudge-Zero, an automated fuzzer designed to identify stealthy input sequences that manipulate the decision-making logic of these models. By exploiting how these models interpret benign formatting symbols and structural tokens, the tool can force an 'allow' decision on otherwise prohibited content. Unlike traditional adversarial attacks that rely on detectable gibberish, AdvJudge-Zero identifies low-perplexity triggers such as markdown headers and specific role indicators. These triggers are virtually invisible to human observers and standard web application firewalls, yet they effectively influence the model's judgments.