Updating Windows Malware Detectors: Balancing Robustness and Regression against Adversarial EXEmples

2024-05-04

Computers & Security

AI Summary

This work addresses the dual challenge of robustness degradation and increased false positives on benign samples (i.e., performance regression) when updating Windows malware detectors under adversarial EXE attacks—formalizing and jointly optimizing this previously unaddressed update trade-off. We propose a dynamic update mechanism integrating gradient-aware adversarial sample filtering with an incremental retraining framework to concurrently enhance robustness and generalization. Our approach synergistically combines PE-specific feature engineering, joint optimization of XGBoost and deep models, adversarial training, and gradient-sensitive sampling. Evaluated on the RealWorld-EXE dataset, our method achieves a 12.7% improvement in adversarial accuracy and reduces the benign false positive rate to 0.3%, significantly outperforming existing detector update strategies.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Computer Vision: Adversarial Attacks & Robustness
• Search and Optimization: Adversarial Search

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Problem

• Balancing robustness and regression in Windows malware detectors
• Detecting adversarial EXEmples without performance regression
• Maintaining accuracy while updating malware detection models

Innovation

• Proposes EXE-scanner plugin for existing detectors
• Balances accuracy and regression in malware detection
• Detects adversarial EXEmples without costly retraining