Intrusion Detection in Fog Computing using Deep Learning Methods

With the enhancement of network throughput, security threats, packet transmission, and attacks in fog computing, Intrusion Detection Systems (IDSs) have been highly utilized to optimally recognize attacks, which are essential for providing adequate security in fog environments. Many researchers have developed diverse deep learning-based approaches for identifying IDSs in fog computing; however, these methods do not detect previously unknown attacks and suffer from high false alarm rates. They are also not capable of dealing with massive amounts of data to recognize security events and their associations. Additionally, they are insufficient for generating adequate security in fog environments.Methodology: Thus, an efficient Federated Learning (FL)-assisted intrusion detection mechanism is developed for identifying threats in a fog computing environment to improve data privacy and prevent data breaches. Additionally, intrusions in fog computing are detected using a Federated-Multi-scale Echo State Network (F-MESN). After detection, the data is encrypted using the Arnold hashing mechanism to protect it from unauthorized access. The encrypted data is securely stored in fog computing for future use. By storing the data safely, the latency while processing the data is minimized. The proposed intrusion detection model is helpful for transmitting sensitive data to the cloud quickly and safely. The performance of the implemented model is analyzed among other techniques to verify its robustness in intrusion detection.Key findings: The simulation outcomes of the developed intrusion detection method achieved higher and optimal results, such as 92% accuracy, 99% specificity, and a 92% NPV measure, demonstrating its superior performance compared to conventional methods.