Cloud-Driven Intrusion Detection Enhancement Using RNN and CNN on Network Traffic
Uman Ahmed Mohammed
Abstract
The intrusion detection systems (IDS) are significant in the protection of networks against malicious activities by tracking and examining the traffic over the network to identify unusual patterns of traffic. With the increasing complexity and bandwidth of network traffic the traditional IDS solutions have serious challenges concerning performance and scalability. This paper proposes an improvement to intrusion detection that is cloud-based, and involves incorporating Recurrent Neural Networks (RNNs) and Convolutional Neural Networks (CNNs) in the context of a cloud infrastructure, thus improving the accuracy and efficiency of traffic analysis. The hybrid architecture proposed takes advantage of the temporal data processing features of RNNs and the strong capacity of CNNs to extract features, which combined, can help to detect complex patterns of intrusion effectively. Empirical assessments have shown that the RNN-CNN composite significantly outperforms the conventional IDS architecture in terms of detection accuracy and a decreased amount of false-positive, as well as the framework is scalable to the cloud environment. The implementation of the cloud computing raises the performance of the model even higher by supplying the necessary number of computational resources to process large volumes of the network traffic in real time. Therefore, the strategy can be viewed as a viable solution to the changing nature of the intrusion detection problem in modern network environments.