Distributed Deep Leaning Attack Detection Framework for Cyber-Attacks in IoT Network
S. Gurusubramani, N. Arockia Rosy, D. Chitra, H. Anwar Basha, Yabesh Abraham Durairaj Isravel, Pundru Chandra Shaker Reddy · 2023
As the frequency of security breaches continues to climb, cyber-security remains a major concern across all industries operating online. Because of the proliferation of new protocols, most of which originate with the Internet-of-Things(IoT), thousands of new zero-day attacks appear every day. Cyberattacks on the IoT have skyrocketed due to the proliferation of connected devices and the inherent security flaws in many network infrastructures. The safety of those systems depends on the ability to recognize and categorize harmful communications. This recommends that even cutting-edge methods, like conventional machine learning (ML) systems, have trouble spotting these atypical, yet dangerous, mutations in attacks over time. However, deep learning's (DL) widespread success in big data applications has piqued the curiosity of many in the cybersecurity community. In order to address multiple potential entry points at once, the authors of this research employ a distributed-framework built on DL to do so. The feedforward neural-network(FFNN) and the long short-term memory (LSTM) are two DL strategies that are compared and contrasted. Both the NSL-KDD and the BoT-IoT datasets are used to test the models' abilities to detect and prevent various cyberattacks. Accuracy of up to 99.98% was achieved across all configurations, proving that the proposed distributed system is successful in detecting multiple classes of cyberattacks.