Optimization of API Aware Malware Detection Using Artificial LSTM Technique

The popularity and open-source nature of APIs have led to a significant increase in malware. Malware developers can evade detection methods, reducing the efficiency of current detection techniques. Therefore, it is crucial for security researchers and experts to devise novel and more effective strategies for analyzing existing and emerging threats on API malware. To effectively examine API security, it is essential to broadly focus on malware development, investigate prevention techniques, and analyze existing detection methods. Traditional malware detection methods are labor-intensive. Machine learning, while instrumental in malware classification and detection, faces challenges in dealing with malware that masquerades as legitimate software and uses self-protection methods, thereby degrading the performance of current machine learning-based methods. This paper presents an analysis of the local malicious behavior of malware and introduces an anti-interference detection framework based on API fragments. The framework leverages an LSTM model to classify API fragments and uses ensemble learning to determine the overall classification of the entire API. We also compare the enhanced accuracy and other parameters of the LSTM-based classification with traditional methods and other machine-learning techniques.