SHAKE-ESDRL-based energy efficient intrusion detection and hashing system

Outstanding progress in unsolicited intrusions along with security threats, which interrupt the normal operations of wireless sensor networks (WSNs), have been attracted by the proliferation of WSNs and their applications. In WSNs, this demands an intrusion detection system (IDS), which can detect such attacks with higher detection accuracy. Designing an effective model for IDS using the SDK-LSHB-based SHAKE-ESDRL algorithm to improve accuracy and lessen training time and response time is the goal of this work. At first, duplicate removal, missing data removal, and data transfer are the steps through which the dataset was processed. From the processed data, by providing the extracted attributes as input to the entropy-based generalized discriminant analysis (E-GDA) method, the number of attributes is reduced. After that, the LogSwish-based deep reinforcement learning algorithm (LS-DRLA) method wielded the reduced attributes for intrusion detection (ID). By utilizing the SHAKE 256 algorithm, the attributes that fall into the attacked class label are hashed and stored in the hash table during this process. Next, to test the real-time data with the trained IDS, the WSN nodes are initialized. For this, by utilizing the supremum distance (SD-K-Means) algorithm, the sensor nodes (SNs) are clustered centered on the cluster heads (CHs) selected by the linear scaling-based honey badger optimization algorithm (LS-HBOA) method. At last, utilizing real-world-based datasets, the proposed algorithms are evaluated and the results are compared using statistical metrics.