Unmasking encryption effects and modified Deep learning approaches for attack classification in WSN.

A widespread application of technology has promoted a rapid development of IoT, where the process of data transfer are made with low latency, higher ranges of effective data transfer using widespread sensors and connections. Nevertheless, the security and the privacy concerns with the WSN are crucially considered for effective means of data transfer using WSN. Thus, the vulnerabilities and the attack against the WSN has to be implemented. With prospects to these advantages of using WSN, many Artificial Intelligence (AI) based concepts such as Machine Learning (ML) models have been used and are established. But the overall outcomes from these models are not satisfactory in terms of their accuracy rates, and are only applicable for small ranges of data. Thus, to overcome these pitfalls, the proposed method has implied an effective Deep Learning (DL) based models for classification. The entire working of the model is adapted in the matlab environment. Initially, the data are generated using the Ant Colony Optimisation (ACO) and the Particle Swarm Optimisation (PSO) for the Cluster Head (CH) selection. The Encryption and decryption approaches for making the data to be secured and are accessible only upon key-generation using Elliptic Curve Cryptosystems (ECC) and Advanced Encryption Standard (AES) algorithm in hybrid mode and are implied with improved rages of feature selection using Genetic Algorithms. For the aspects of classification of attack and non-attack upon WSN, the combination of Modified Deep-Convolutional Neural Network-Bi-Directional Long Short Term Memory with Attention Mechanism is established (MDCNN- Bi-LSTM- AM). The effective working of the proposed model performing the classification of attack and non-attack upon WSN is evaluated using applicable and measurable metrics. Further affirmative ability of the proposed model is carried out by comparison with state-of-art approaches. [ABSTRACT FROM AUTHOR]