A method for malware detection based on CNN-BiLSTM

With the rapid advancement of computer technology, cybersecurity incidents are occurring frequently, and malicious activities on the Internet are on the rise, posing a serious threat to the security of information and data for nations, enterprises, and individuals. Malware detection stands as a pivotal solution in upholding cybersecurity, holding significant practical significance in maintaining network order and safeguarding public information integrity. This paper centers on the issue of privacy data exposure that may arise from the abundance of traffic features required for training malware detection models. By delving into the spatiotemporal characteristics of traffic data and integrating deep neural networks with federated learning, a malicious software detection solution based on the CNN-BiLSTM model is established in the context of distributed traffic data. Comparative experiments are conducted with detection models based on random forests and autoencoders, demonstrating that this model exhibits commendable detection accuracy in distributed scenarios.