Optimization of multilayer perceptron architecture using particle swarm optimization for enhanced network attack detection on IoT devices

The study conducts an analysis of recent research of the use of neural networks in conjunction with the Particle Swarm Optimization (PSO) algorithm for detecting anomalies in IoT device network traffic. The aim of the study is to investigate the potential of the Particle Swarm Optimization algorithm for the automated design of neural network architecture. Materials and methods: an analysis of scientific literature was carried out, and a comparative experiment was conducted for two Multilayer Perceptron (MLP) models. The architecture of one model was designed using Particle Swarm Optimization, while the other was designed empirically. The experiment was conducted for the task of classifying IoT device network traffic. The following results were obtained during the study: it was experimentally confirmed that the MLP architecture optimized using the PSO algorithm outperforms the empirically designed MLP architecture in terms of classification task quality. In conclusion, it is noted that the method of automated neural network architecture design based of the PSO algorithm demonstrated its effectiveness, surpassing the empirical approach. The work contributes to the development of the Neural Architecture Search (NAS) field, opening prospects for creating more efficient IoT security models and applying PSO for optimizing architectures of other types of neural networks.

Authors: D. S. Dobrovolskii, Yu. N. Kozhubaev

Direction: Informatics, Computer Technologies And Control

Keywords: particle swarm algorithm, neural networks, multilayer perceptron, Internet of things, optimization, anomaly detection, neural architecture search

---

View full article