Intrusion Detection System Attack Detection and Classification Model with Feed-Forward LSTM Gate in Conventional Dataset

Abstract

With the increased network scale, intrusion detection is more frequent, advanced, and volatile for capturing large scale is challenging. The increased malicious attacks in the network system affect the security tool in the network causing illegal users, reliability, and robustness for network security. Recently, network security is increased with the illegal intrusion detection system for the security in each occasion for the sensitive users, governments, enterprises and governments. Network Intrusion Detection (NIDS) exhibits a reliable and effective technological form for the packets in the network, unauthorized users, and traffic monitoring for the computer network. The incorporation of the machine learning model exhibits effective performance for network traffic monitoring. Additionally, the NIDS model exhibits effective attack detection and traffic malicious activities in the network. Through intelligent capability, the machine learning model comprises of intrusion detection based on rule-based solution for the network attacks. This paper proposed a Long Short Term Memory Gate Recurrent Neural Network (LSTMgateRNN). The constructed LSTMgateRNN model comprises of the LSTM features for the attribute evaluation for the attack data processing. The incorporation of the gate function within the network the incorporated gate function increases the evaluation of the attributes in the network. Finally, the processed data is examined with the RNN model to perform the attack detection and classification. The proposed LSTMgateRNN model performance is evaluated for the three different datasets such as KDD’99, NSL-KDD and UNSW-SW15. Simulation analysis exhibited that proposed LSTMgateRNN model achieves an attack detection rate of 99%. The proposed LSTMgateRNN is comparatively examined with the with the existing model and achieves the ~6 – 12% improved performance.