Llm malware analyzers tricked by false security stories in binaries
ALIBI: Adversarial Legitimacy Injection in Binary Input against LLM Malware Analyzers
Cryptography and SecurityMachine Learning
Summary
Malware detectors that use large language models (LLMs) to analyze software can be fooled by adding a fake but believable story inside the program. This trick makes the detector think the program is safe, even though it is harmful. The researchers showed how adding a small hidden text section can change the analyzer’s decision without changing the program’s actual behavior. They also tested ways to reduce this trick, but found it is still hard to rely on these AI-based malware checkers without verifying the true facts separately.
What this means in practice
- •For security teams: Detect and defend against semantic cover story attacks when using LLM-based malware analyzers in threat detection workflows.
- •For software developers: Design software inspection tools with provenance checks that separate verified code facts from attacker-controlled text narratives in binaries.
Authors
Hyeongjun Choi, Wonyoung Jung, Haehoon Seo, Sungyup Nam
Abstract
Large language models are being integrated into malware triage workflows as reasoning components that summarize static evidence and produce analyst-facing verdicts. This paper shows that the same reasoning capability introduces a new attack surface. We present ALIBI, a semantic cover story attack against frontier LLM-based malware analyzers. ALIBI adds a small, non-executed read-only section to a compiled binary, containing a coherent but false security product narrative, without altering imports or executable behavior. Instead of issuing direct instructions to the model, it reframes suspicious evidence as expected behavior of a benign endpoint security tool. On a frozen PE set of 50 malicious samples, the payload flips 30 of the 35 baseline-malicious samples to benign on Gemini 2.5 Pro, while GPT-5.5 Pro and Claude Opus 4.7 produce substantial severity downgrades with significant confidence reductions even when verdict labels are preserved. The attack transfers to ELF binaries, where Gemini flips 16 of 40. A verification-guided defense prompt roughly halves the benign verdicts, but 42.9 percent of malicious samples still reach benign. LLM malware analyzers therefore require provenance checks that separate verified facts from attacker-controlled claims, not narrative trust.