Enhanced IoT Security for DDOS Attack Detection: Split Attention-Based ResNeXt-GRU Ensembler Approach

Abstract

The rising Internet of Things (IoT) device count has caused security concerns among high-tech companies and groups, which has resulted in several evaluations. IoT’s pervasive, portable, and intelligent qualities make developing automated methods for spotting suspicious activity on IoT devices linked to regional infrastructure very vital. Using input factors, including network traffic attributes and output parameters, including accuracy, time complexity, and specificity, this work examines datasets, including NSL-KDD, CIC-IDS17, ToN_IoT, and UNSW-NB15. Our suggested approach uses a ResNeSt model with Split-Attention (ResNeSt), improved using the Jaya Algorithm (RSG-MJ) and augmented by a Gated Recurrent Unit (GRU). This method attained a 15% boost in efficiency considering computational complexity and an accuracy rating of 98.45%. Early-stage threat detection and computing efficiency of our system show significant advances over current techniques. The statistical analysis measures support the resilience and efficiency of our approach even more in line with the journal’s goal of advancing IoT security via creative approaches. Our method is a viable solution for real-world IoT security issues as the testing results reveal its faster detection of DDoS attacks, hence improving performance.

Keywords

• DDOS attack
• IoT
• intrusion detection
• deep neural networks
• optimization techniques
• security