VISUAL IMAGE-BASED MALWARE DETECTION SYSTEM USING DEEP LEARNING

• Main Author: Ainun Rofik, Muh.
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/79488
• Institution: Institut Teknologi Bandung

The escalating strength and complexity of malware threats surpass the capacities of conventional defense systems. Traditional static analysis prevention, although highly accurate in detecting well-known malware, struggles against polymorphism and packaging evasion techniques. Dynamic analysis, on the other hand, often yields false positive results at a high frequency. A prospective avenue in the evolution of malware detection involves visual image-based identification. This study pioneers the integration of image-based representation and deep learning for classification to discern the presence of malware. Not only does this approach successfully identify malware, but it also excels in classifying malware based on its family. The test outcomes demonstrate an impressive accuracy of 98.88% in malware detection and 94.40% in malware family classification. This research contributes to the ongoing discourse on enhancing cybersecurity measures by leveraging innovative techniques to combat the relentless evolution of malicious software. As the digital landscape evolves, adopting image-based identification methods offers a promising trajectory for robust and effective malware detection and classification.