Self-Supervised Learning Implementation for Malware Detection

The huge and ever-increasing amount of malware complicates the malware detection process. To detect malware, antivirus still relies on signature-based and heuristic-based detection techniques. However, malware signatures and heuristics rules are compiled by malware analysts manually. It takes a long time and special skills to analyze malware and create a malware signature. To tackle this problem, machine learning is implemented for malware detection. However, the implementation of machine learning on malware detection is still experiencing several problems. It requires a large dataset, and the dataset labeling process takes a long time. For this reason, a new malware detection method approach is needed. The self-supervised learning method does not require large datasets and manual labeling processes. Self-supervised pretraining a model on unlabeled data and then fine-tuning it on a smaller dataset Self-supervised learning has resulted in good performance in NLP and computer vision. In this paper, a study was conducted on the self-supervised learning method and its possible implementation for malware detection. A malware detection experiment with self-supervised learning was also conducted using BERT. The experimental results show a promising accuracy of 85%.