LLMs enhance malware detection by analyzing behavioral reports, outperforming static methods.

Researchers have developed Trident, a new system that enhances malware detection by integrating large language models (LLMs) with behavioral analysis. Unlike traditional methods relying on static features, Trident processes semi-structured sandbox behavior reports using LLMs to generate robust, concept-drift-resistant detection rules. The system combines these LLM-derived rules with a classic decision tree model and direct LLM analysis of sandbox outputs, outperforming existing static-feature and behavior-based approaches. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Enhances malware detection robustness against concept drift by leveraging LLMs for behavioral analysis.

RANK_REASON Academic paper introducing a new system for malware detection using LLMs and behavioral features.

Read on arXiv cs.LG →

• Trident
• LLMs
• arXiv
• PE malware

• paper
• safety