TRACKING MALICIOUS SOFTWARE THROUGH LOG-BASED BEHAVIOURAL SIGNATURES

Cyberattacks, particularly malware, have evolved significantly, making traditional signature-based detection methods ineffective against modern threats such as zero-day exploits and polymorphic malware. This project presents a machine learning-based framework for malware detection, leveraging application logs and behavioral analysis to identify malicious activities accurately. By extracting critical system features—including resource utilization, file access patterns, and network anomalies—the system creates a robust dataset for training classification models such as Random Forest (RF) and Support Vector Machine (SVM). Unlike conventional antivirus solutions that rely on predefined signatures, this approach dynamically learns patterns from historical and real-time data, enabling proactive identification of threats. The integration of advanced machine learning techniques ensures a high detection rate while minimizing false positives, making it a scalable and adaptive solution for evolving cybersecurity challenges. The model’s ability to analyze large-scale system logs in real- time enhances malware detection efficiency and provides an additional layer of defense against sophisticated attacks. The proposed system offers significant improvements in speed, accuracy, and adaptability over traditional detection methods. By utilizing machine learning, the framework not only detects known malware but also effectively identifies novel and obfuscated threats, contributing to a more resilient cybersecurity infrastructure.