IoT Anomaly Detection Using Picture Fuzzy Clustering Approach

Abstract

Improving the effectiveness of security systems without slowing them down is a major challenge in cybersecurity. Many methods have been explored for detecting anomalous behaviour in network data, with fuzzy set-based approaches standing out for their potential. The Internet of Things (IoT) consists of many connected devices that constantly generate large amounts of data and perform tasks in real time.  Because they are always online, these devices are especially vulnerable to cyberattacks. Detecting such malicious activity considered anomalies in the data is a key research issue. Picture Fuzzy Sets (PFSs) offer a robust means to deal with the uncertainty, vagueness, and imprecision found in IoT data. PFSs are constructed on Intuitionistic Fuzzy Sets (IFSs) by introducing a neutrality component, in addition to membership and non-membership values. In this paper, we propose a method based on Picture Fuzzy C-Means (PFCM) clustering to detect anomalies in IoT data. This method is an improved version of the traditional Fuzzy C-Means (FCM) algorithm and is better suited to handling the multifaceted uncertainty in IoT environments. We also evaluate the computational efficacy through complexity analysis. Experimental results using real-world datasets, NSL-KDD and SAB, show that the detection rate and accuracy improve by up to 10% compared to the k-means algorithm, 5% compared to FCM, and 3% compared to the Intuitionistic Fuzzy C-Means (IFCM) algorithm, demonstrating the superior efficacy of our approach.