BERT‑LSTM‑LGBM Approach for DDoS Attacks Detection in IoT Network Using ML

Abstract

New cybersecurity challenges have increased as the interconnected IoT devices grow, such as DDoS attacks, which are observed as more attacks exploit resource‑constrained IoT devices. Conventional detection mechanisms often fail to capture the dynamic and diverse nature of IoT network traffic, and several researchers and professionals have addressed these concerns. In view of the issues raised by the researchers, the presented models need to enhance their accuracy and performance. The BERT_LSTM‑LGBM model has been proposed for an intelligent and accurate DDoS attack detection in IoT devices. BERT component is used to remove deep contextual features from network traffic data, capturing intractable relationships and semantic dependency. The long Short‑Term Memory (LSTM) network further improves temporal arrangements learning to detect sequential anomalies, while the LGBM classifier promises high‑speed and comprehensible decision‑making. The results show that the BERT‑LSTM‑LGBM framework is robust and can detect diverse DDoS attack patterns, offering a scalable and intelligent solution for securing next‑generation IoT infrastructures. Our proposed model presents its exceptional proficiency in threat detection within the IoT environment. We achieved remarkable results such as 99.8%, 98%, and 99%.