Reinforcement Learning for Intrusion Detection and Improving Optimal Route by Cuckoo Search in WSN K.Sai Madhuri Research Scholar, Department of Information Science and Engineering, Visvesvaraya Technological University, Belagavi. Nagarjuna College of Engineering and Technology, Research Centre, Bangalore, Karnataka, India saimadhuri069@gmail.com Dr. Jitendranath Mungara Principal, Department of Computer Science and Engineering, Nagarjuna College of Engineering and Technology, Bangalore, Karnataka, India Abstract Wireless Sensor Network (WSN) is a generally hopeful technology for several real-time applications due to its cost-effective, size, and distribution nature. WSN is a collection of sensor nodes spread in a great region such that the required information can be collected. However, sensor nodes are susceptible to attacks, for example, intrusion, hackers, defective hardware starting the physical incident, etc. Therefore, it is compulsory to defend a sensor node from an intrusion. If it brings attacked next, the information transmitted through the sensor may be wrong and lead to incorrect data analysis, leading to unnecessary outcomes. To solve these issues, Reinforcement Learning for Intrusion Detection (RLID) and Improving Optimal Route by Cuckoo Search is proposed. The Reinforcement Learning uses the repeating node classification for detecting the intrusion during the route discovery. Reinforcement learning evaluates the sensor node behaviour by the quality of the link, and it is computed by sensor node packet forward rate and node residual energy. Here, the repeating node classification method classified the intrusion sensor based on node-link quality. As a result, it can improve intrusion detection performance efficiently. Besides, the Cuckoo Search Technique (CST) is used to find the optimal forwarder for transmitting the data from sender to destination. The main objective of this work is to offer optimal routing and communicate the data via normal sensor nodes in WSN. The simulation platform and the obtained results are compared with the baseline protocol to prove the efficiency of our proposed approach. Keywords: Wireless sensor network, Repeating node classification, Reinforcement Learning, Cuckoo search technique, Intrusion Detection. 1. Introduction WSNs attained fame as they can adjust to the updates in a physical setting, for example, pressure, temperature, sound, and pollution. The advantage of such structures is that they are flexible, suitable for isolated places such as mountain areas, seas, forests, and rural areas. WSNs have been extensively useful in various areas, for example, surroundings supervising, political, intelligent transportation military, industrial fields and agricultural also medical (sen et al. 2018). In WSN, intrusion denotes the trouble of observing and separating flows and performance from the usual behavior that can unfavourably crash the information security (Sun et al. 2015). Owing to the enormous development of Internet applications, the necessity for data security has enlarged multiple. Because a major defence of network structure, a Detection of Intrusion is accepted to adjust to energetically altering risk scenery. Cryptographic key management is a complex system, and it is an expensive process. Pre-configuring plus cryptographic keying material is a precondition for transition link securing if indirect key validation is not accessible. This approach provides security to every intermediate node through authentication using Dij-Huff Approach (DHA). In this approach, the Huffman coding offers security, and all intermediate nodes provide security using the Binary Hop Count method. However, the Binary Hop Count method does not work better during a malevolent attack (Alghamdi et al. 2018). Trust management is measuring trust with properties which manipulate trust. Bayesian-based Trust Management Approach (BTMA) is used for detecting the intrusion. However, trust management creates several problems, for example, limitation of necessary valuation data, require of big data procedure, the demand of easy trust correlation appearance as well as the expectation of automation (Meng et al. 2017). e-ISSN : 0976-5166 p-ISSN : 2231-3850 K.Sai Madhuri et al. / Indian Journal of Computer Science and Engineering (IJCSE) DOI : 10.21817/indjcse/2021/v12i6/211206024 Vol. 12 No. 6 Nov-Dec 2021 1760