A highly secured intrusion detection system for IoT using EXPSO-STFA feature selection for LAANN to detect attacks

The Internet of Things (IoT) is a modern age technology, designed with the vision to connect and also interconnect all the objects everywhere. Technological progressions provide businesses with many comforts as well as helps the attackers and intruders to crack the IoT networks’ security. Numerous Intrusion Detection Systems (IDSs) are created aimed to attack prevention systems. Frequently, security stays to be challenging in the IoT networks. The work addressed here presents the new effective secured IDS aimed at IoT environment, which sustains the data’s confidentiality, integrity, together with its availability. At first, the data has been pre-processed, which helps in acquiring a clear vision about any attack that is about to occur. The methods are handling of missing and Nan values, date and time variables, categorical features and with scaling of data. Next, aimed at acquiring the data’s knowledge, this work has established an Improved Pearson Correlation Coefficient (IPCC), Feature Extraction (FE) method that presents the relation amidst the data by pondering the causative. The features’ extraction is next followed by the relevant features’ selection aimed at maintaining an efficient computational time and also accuracy utilizing Explorated Particle Swarm Optimization (PSO) centred Sea Turtle Foraging Algorithm (EXPSO-STFA). At last, the feature chosen has been trained and then examined over the Look Ahead Artificial Neural Network (LAANN) classification aimed at identifying the attacks. The LAANN method offers lesser error rate and also evades False Alarm Rate’s (FAR’s) chances and also locates the attack much effectively and also reliably. Moreover, the work administers the malicious attacks’ and random behaviour, and also yields an accurate outcome with the help of evaluation parameters such as Accuracy, Specificity, Sensitivity, Precision, F-Measures, FPR, FNR and MCC. Experiential examination exhibits that the work yields 95.65% accuracy, and also attains 98.16% average Attack Detection Rate (ADR), and the work stays to be much scalable and also secured analogized to the existent top-notch techniques.