Network Intrusion Detection and Mitigation Using Hybrid Optimization Integrated Deep Q Network

Abstract

Secure communication between network resources is a key issue, and researchers are persistently developing new intrusion detection schemes to detect intruders to enhance security. However, the traditional intrusion detection techniques have security challenges due to their poor detection performance. Hence, an effective intrusion detection model, namely spider monkey social optimization algorithm (SMSOA)-based Deep Q network is introduced for the detection of intrusion in the network. Here, the intrusion is detected based on the following three steps, such as pre-processing, feature fusion step, and detection. Estimated value replaced the missed value using missing value imputation in the pre-processing phase. Jaccard index by means of deep belief network (DBN) is performed in feature fusion. Deep Q network is used for intrusion detection process, where SMSOA method is used for training. The developed SMSOA scheme is performed by the integration of the social optimization algorithm (SOA) and spider monkey optimization (SMO). In addition, the attack mitigation process prevents the intruder enter into the network. Moreover, the developed model attained a better performance with respect to f-measure, recall values, and precision of 0.9455, 0.9585, and 0.9590, respectively.

Keywords:

• Deep belief network
• Deep Q network
• Jaccard index
• social optimization algorithm
• spider monkey optimization

Data Availability Statement

The data underlying this article are available in BOT-IOT dataset, at “https://ieee-dataport.org/documents/bot-iot-dataset”.

The data underlying this article are available in NSL-KDD dataset, at “https://www.unb.ca/cic/datasets/nsl.html”.