Enhanced Security: Detection of Android Malware Using Siamese Bidirectional LSTM

Abstract

Advanced detection techniques are required due to the serious threat that Android malware poses to mobile security. Conventional signature-based methods frequently fall short in detecting novel, changing threats. This study introduces a novel method for detecting Android malware that makes use of machine learning and the Siamese Shot Learning (SSL) technique. Large datasets of both malicious and benign applications are used to train machine learning models, which then extract features like network behavior, permissions, and API calls. To improve detection accuracy, especially for new malware variants with little training data, the SSL technique is used. SSL makes the model extremely effective against zero-day attacks by utilizing few-shot learning, which allows the model to identify malicious patterns even with few examples. The experimental findings show that the suggested approach performs better than traditional classification models in terms of overall detection rate, recall, and precision. The detection framework's robustness is greatly increased when feature engineering, deep learning, and SSL are combined. By offering a scalable and effective method for detecting Android malware and lowering dependency on heavily labeled datasets, this research advances the field of cybersecurity. By proactively identifying threats with high accuracy and adaptability, the suggested system improves mobile security. To further improve mobile device protection, future research will concentrate on improving feature selection and incorporating real-time detection mechanisms. These results highlight how machine learning and SSL can protect Android ecosystems from increasingly complex cyberthreats, providing a promising path for automated and intelligent malware detection.