An Efficient DDoS Attack Detection Using Optimized Long Short-Term Optimization Based on Improved Brainstorm Optimization

Objectives: The primary result of the present research will be the creation of an effective and powerful deep learning architecture that can be utilized to correctly identify DDoS attacks in IoT networks through optimal hyperparameter adjustment of LSTM via an Improved Brain Storm Optimization (IBSO) algorithm combined with Centroid Opposition-Based Learning (COBL). It is also the purpose of the work to examine how various combinations of activation functions influence the performance of detection and to demonstrate quicker convergence to better generalization of realistic IoT traffic. Methods: A LSTM model is suggested, which has been optimized using the IBSO, and in which COBL is considered as an addition to the Brain Storm Optimization algorithm in order to increase the diversity of the population, improve global exploration, and prevent early convergence. The IBSO algorithm is used to automatically optimize such key hyperparameters of LSTM learning rate, number of hidden neurons, batch size, and activation functions. Training and validation are done using the CIC-IoT-2023 dataset that consists of real-world IoT traffic and various scenarios of DDoS attacks. The efficiency of performance is measured in terms of accuracy, MSE, and convergence, and is compared to the currently available deep learning-based DDoS detection models. Findings: The suggested IBSO+LSTM model has a high detection rate of 98.92 and is better than a few other state-of-the-art models. The convergence analysis, which depends on accuracy and MSE, shows that the process of optimization with IBSO results in faster and more stable learning than the traditional tuning and standard optimization methods. The research also shows that there are combinations of activation functions that are highly effective in enhancing the classification performance and strength. Novelty: This work is novel in that it combines COBL and BSO in tuning hyperparameters to LSTM to detect IoT DDoS. This IBSO hybrid strategy is a good balance between exploration and exploitation, convergence and local optima. Furthermore, a systematic analysis of activation function combinations and testing on the recent CIC-IoT-2023 data offers a well-rounded and realistic structure of high-accuracy, scalable, and resource-efficient DDoS detection in the real-world IoT setting. Keywords: DDoS attack detection, Internet of Things, LSTM, Improved Brain Storm Optimization, Centroid Opposition-Based Learning, CIC-IoT-2023