Advanced Framework for Detecting Malware in Portable Executable (PE) Files Using a Multi-Model

Abstract

This paper presents Malware detection in Portable Executable (PE) files remains a critical challenge in cybersecurity, with attackers increasingly using obfuscation, polymorphism, and zero-day exploits to evade detection. Malware has emerged as a major problem in today's digital era. The malware goals are to interfere with, damage, or compromise information system and computer system without the operator's approval or knowledge. At present, malware is considered among the most common cyber threat We combine static, dynamic, and hybrid analysis techniques with ensemble learning to achieve superior detection accuracy compared to traditional methods. Our system integrates feature engineering from PE headers, byte-level CNN analysis, API sequence modeling with LSTMs, and attention mechanisms via Transformers, culminating in a stacked ensemble classifier. Experimental results demonstrate 98.7% detection accuracy with only 0.8% false positives on a dataset of 100,000 PE files. The paper provides complete mathematical formulations, architectural details, and empirical validation of our approach.