IoT Security Detection and Evaluation for Smart Cyber Infrastructures Using LSTMs with Attention Mechanism

Abstract

A computer network may be impacted by malware, computer viruses, and other hostile attacks. One of the most important elements of network security as an active defence technology is Intrusion Detection System (IDS). Issues with traditional intrusion detection systems include low detection effectiveness, inaccurate results, and an inability to handle new types of intrusions. This paper implemented a deep LSTMs with attention mechanism approach for IDS. The proposed technique is determined on Bot-IoT dataset, and z-score normalization is performed to normalize data. Information Gain (IG) is utilized to choose the features. A deep LSTMs with attention mechanism is used to effectively identify and categorize IDS effectively. The IG-deep LSTMs with attention mechanism achieves better accuracy using of 99.70% and 99.60% in NSL-KDD and CIC-IDS2017 datasets respectively.