Detection of Malicious Interpreted Code using ML to Optimize Parameters for Minimizing Prediction Errors

Mobile devices face growing cybersecurity threats, including malware, data breaches, malicious applications, and social engineering attacks. Cybercriminals exploit these vulnerabilities to gain unauthorized access, disrupt communication, and compromise sensitive information. Conventional malware detection methods, such as signature-based and heuristic techniques, often struggle to identify evolving threats due to their reliance on predefined patterns. This study presents an advanced malware detection framework that integrates a Reservoir Network with a Deep Learning Approach (DLA) to enhance detection accuracy and minimize prediction errors. The proposed method utilizes two static features-API call permissions and alert signals- to improve detection efficiency. Experimental evaluation, conducted using the Drebin malware dataset, demonstrates that the proposed model achieves an accuracy of 99.7 %, reduces the false positive rate to below 1 %, and ensures optimal processing speed with a batch size of 64 and a learning rate of 0.001. Compared to traditional approaches, this framework exhibits superior adaptability to obfuscated malware while enabling real-time threat detection. The findings emphasize the effectiveness of machine learning (ML)-driven security mechanisms in addressing emerging cybersecurity threats.