Abstract

The development of mobile malware requires adaptive intelligent analysers to change with the new threats. The current paper features an AI-based Android APK forensic system that uses a hybrid deep learning model that consists of the BERT, CNN, and LSTM models to automatically infer transformation plans using security bulletins. The system uses ten methods of transformation of encryption, obfuscation, and signature modification forms. The experimental assessment of 6,000 samples of security data shows 92.00%, 90.01%, and 87.99% accuracy, precision, and recall, respectively. The framework, which is deployed through the interface of a Telegram bot, can be used in security research, forensic investigation, and antivirus testing, with the ability to document all the results thoroughly.