A Data-Driven Approach to IoT Security: Detecting Cyber Attacks with AEInc-BGTO

Abstract

Cyberattacks on the Internet of Things have significantly raised due to the extensive adoption of smart devices and the multiple security flaws in networks. It is essential to identify and categorize malicious traffic to guarantee the security of these systems. Nonetheless, the earlier research has encountered obstacles such as limited characteristics, reduced effectiveness, irrelevant attributes, and complex computational tasks. This research proposes an innovative framework called the Attention Embedded Inception-based Bidirectional Gorilla Troop Optimization strategy, which is an emerging mechanism for detecting and classifying various kinds of attacks to ensure security in Internet of Things settings. Some preprocessing procedures are performed to improve the quality of received data and enhance information security. The feature extraction process is performed by Convolutional Block Attention Embedded InceptionV4, which is developed by integrating the Convolutional Block Attention Mechanism with InceptionV4. Moreover, the suggested Bidirectional Long Short Term Memory approach effectively categorizes and detects several types of attacks in IoT, and Gorilla Troops Optimization technique is implemented for fine-tuning model’s hyperparameters, thereby enhancing detection accuracy, and security rate. The proposed strategy is compared to previous attack detection mechanisms in terms of some common assessing metrics using five attack datasets, where it is obtained that the proposed strategy yields commendable findings. The proposed strategy achieves detection accuracy of 99.27% and security rate of 99.18%, which outperforms earlier assault detection techniques. Consequently, it is claimed that the proposed strategy is the most effective for several attack identification on secure Internet of Things.