Unveiling Malicious Patterns: Autoencoder-Based Malware Detection

Malware detection poses a significant challenge in cybersecurity, particularly with the increasing sophistication of attack methods. This study introduces an autoencoder-based approach to detect malware by learning the structure of benign data and identifying anomalies through reconstruction loss. By focusing on the detection of deviations in data patterns, this method offers an effective solution for identifying both known and unknown malware. Using the MALIMG dataset, the approach is evaluated with standard metrics such as accuracy, precision, recall, and F1-score, demonstrating strong performance and computational efficiency. This work highlights the potential of autoencoders as a robust anomaly-based detection tool.