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

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 →

• arXiv
• LLMs
• PE malware
• Trident

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →