Malicious Data Classification in Packet Data Network Through Hybrid Meta Deep Learning

Advancements in wireless network technology have provided a powerful tool to boost productivity and serve as a vital communication method that overcomes the limitations of wired networks. However, because of using wireless networks, security is an increasing concern in the community. At the time of our study, people rely on machine learning techniques to create a trustworthy networking system. However, it hinders the development of a reliable network as the number of publicly available malicious data is insufficient to train a model correctly. In real life, people are not very keen to share this data as they are sensitive. In order to deal with this issue, we primarily aim to develop a solution that provides a reliable intrusion detection system despite being trained with a small amount of data. This paper proposes a novel idea of hybrid meta deep learning in detecting malicious packet data. We use a combination of Siamese and Prototypical networks where the Siamese network is used for binary classification and the Prototypical network for multi-class classification. Both approaches are based on meta learning techniques, requiring a minimal amount of data for most attack classes. Utilizing these meta learning characteristics, we could train our model with just 3000 data samples and achieve more than 90% accuracy for both meta learning tactics. Our study aims to provide a secure and trustworthy network domain that enhances communication between end users.