LIoTID: LSTM based algorithm for IoT botnet intrusion detection
Vikrant, Gesu Thakur · IET conference proceedings. · 2025
The Internet of Things (IoT) has grown significantly since its conception, marking a game-changing technical innovation. In essence, the IoT network is the integration of network and sensors to automate and centralize many activities. This transformational technology is disrupting company processes and transforming society. As IoT evolves, the significance of detecting vulnerabilities and flaws that may decrease the unauthorized access and prevent the crucial resources, which might potentially render the entire system inoperable. One common threat in IoT is Denial of Service (DoS) and Distributed DoS threats. In the current article, a novel architecture called LSTM-based IoT intrusion detection (LIoTID). The dataset for IoT botnet is built by collecting packets from IoT traffic and UNSW-NB15 and Bot-IoT dataset feature comparison. It also focuses on the properties of the Transmission Control Protocol (TCP) protocol. The main objective is to correctly categorize network traffic into three groups: DoS, DDoS, and non-anomalous, while taking overfitting and data imbalance into account. By applying deep learning (DL) methods such as LSTM, and able to achieve a remarkable 96.3% classification accuracy. This degree of precision shows how well the proposed system, called LIoTID can detect and stop infiltration attempts in IoT network, improving the security and dependability of IoT systems.