New benchmark EXE-Bench ranks AI malware detectors on usability

Researchers have developed EXE-Bench, a new benchmark designed to comprehensively evaluate AI-based Windows malware detectors. Existing evaluations often lack systematic comparison, differing in data, temporal analysis, adversarial robustness, and computational requirements. EXE-Bench addresses these gaps by assessing performance, temporal and adversarial robustness, and computational overhead, consolidating them into a single score for fair model comparison. The benchmark highlights the continued utility of domain knowledge through feature engineering, which demonstrates resilience against time and adversarial attacks, contrasting with deep neural networks that may only perform well immediately after deployment. AI

IMPACT Provides a standardized method for evaluating AI malware detectors, potentially improving real-world deployment and security.

RANK_REASON Research paper introducing a new benchmark for evaluating AI-based malware detectors. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Deep Neural Networks
• EXE-Bench
• Hugging Face
• malware
• Microsoft Windows

• paper
• safety

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