Multi-Label Classification for IoT Attack Detection Using Adaptive Neuro-Fuzzy Inference System

Vilchez V.; Hinojosa E.; De Grande R.E.; Astudillo C.A. · Proceedings - 2025 21st International Conference on Distributed Computing in Smart Systems and the Internet of Things, DCOSS-IoT 2025 (2025) · DOI: 10.1109/dcoss-iot65416.2025.00098

Resumen

Multi-label classification (MLC), essential in domains like the Internet of Things (IoT), assigns multiple nonexclusive labels to each input, posing challenges such as managing label dependencies, handling large-scale data, and adapting to dynamic environments. Traditional methods struggle with these complexities, leading to low accuracy and limited interpretability. This work introduces an Adaptive Neuro-Fuzzy Inference System (ANFIS) for MLC, designed to enhance accuracy and handle label correlations, particularly in detecting Distributed Denial of Service (DDoS) attacks. By leveraging shared antecedent parameters and independent consequent parameters, ANFIS improves classification performance and adaptability. Evaluations on a synthetic dataset demonstrate its superiority over traditional algorithms, achieving higher Average Precision and Area Under the Curve scores, making it a strong candidate for real-time deployment in large-scale IoT security applications.

Instituciones: Universidad Nacional de San Agustín de Arequipa