Leveraging Machine Learning for Network Intrusion Detection in Social Internet of Things (SIoT) Systems - Research Abstract

Aug 3, 2026

Research explores machine learning–based network intrusion detection for Social Internet of Things (SIoT) systems, focusing on identifying anomalous behavior in socially interconnected environments. CNN+GAN, GAN, and Logistic Regression are evaluated using the CIC IoT Dataset 2023 to detect diverse intrusion types. CNN+GAN delivers the most accurate results, indicating stronger capability for complex threat patterns. Findings guide selection of intrusion detection methods and support ongoing work to handle evolving attacks and SIoT vulnerabilities.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Abstract

Introduction

Challenges and Vulnerabilities

Motivation

Leveraging Machine Learning for Network Intrusion Detection in Social Internet Of Things (SIoT) Systems Divya S1, Tanuja R2 1,2 Department ofCSE UVCE, Bengaluru, India [divyasdivu1994@gmail.com](divyasdivu1994@gmail.com), [tanujr.uvce@gmail.com](tanujr.uvce@gmail.com) Abstract—This research investigates the application of machine learning models for network intrusion detection in the context of Social Internet of Things (SIoT) systems. We evaluate Convolutional Neural Network with Generative Adversarial Network (CNN+GAN), Generative Adversarial Network (GAN), and Logistic Regression models using the CIC IoT Dataset 2023. CNN+GAN emerges as a promising approach, exhibiting superior performance in accurately identifying diverse intrusion types. Our study emphasizes the significance of advanced machine learning techniques in enhancing SIoT security by effectively detecting anomalous behaviours within socially interconnected environments. The findings provide practical insights for selecting suitable intrusion detection methods and highlight the need for ongoing research to address evolving intrusion scenarios and vulnerabilities in SIoT ecosystems. Keywords-SIoT Security, Intrusion Detection, AI/ML. CNN, GAN. I. INTRODUCTION In contemporary society, the Internet of Things (IoT) has emerged as a transformative force across diverse industries. With applications spanning healthcare, transportation, and beyond, IoT's interconnected sensor networks generate substantial network traffic. This paradigm shift has ushered inan era of increased IoT integration into daily life [1] . Notably, IoT technology has revolutionized healthcare by enabling continuous patient monitoring [2],[3], and in transportation, it helps accident detection [4],[5] . Industrial IoT (IIoT) has introduced reliable, low-latency monitoring and control solutions [6] . IoT's impact has extended to education, aviation, forestry, and more [7],[8] . IoT connections have surged, promising continued growth [9],[10], and propelling innovative business models and distributed infrastructure concepts. However, formidable challenges persist, encompassing interoperability, security, and standardization [11],[12] and [13] . Unique applications like Internet of Vehicles (IoV) demand stringent response times [14] . Detecting attacks on IoT devices remains complex due to distributed connectionsand security gaps [15],[16] and [17] . Despite efforts to create attack datasets, gaps remain. Many attacks go unrepresented, and real-world IoT device networks are often overlooked. Additionally, the need for datasets featuring malicious IoT devices executing attacks is evident. To develop effective security analytics for intrusion detection, comprehensive data is essential, encompassing diverse attack types, real IoT device networks, and malicious IoT deviceexecuted attacks. Figure 1. Depicting the SIoT relationships Figure 1 illustrates the intricate relationships within the Social Internet of Things (SIoT) ecosystem, categorizing them into five distinct levels: At the core of the figure, there is a large circle labelled "Social Level." This represents the overarching social dimension of SIoT, emphasizing interactions and relationships among users. Surrounding the "Social Level" circle, there is a concentric circle labelled "Physical Level." This outer circle signifies the physical aspects of SIoT, encompassing the devices and objects that form the foundation of the ecosystem. Arrows extend from the "Social Level" circle to the "Physical Level"circle, symbolizing the relationships between users and IoT objects. These relationships capture how users interact with and control IoT devices in their environment. Within the"Social Level" circle, there are additional arrows connecting users to one another. These arrows signify the interpersonal connections and collaborations among users within the SIoT ecosystem. These relationships may involve communication, data

FAQs

Which machine learning models are evaluated for SIoT network intrusion detection?

The study evaluates CNN+GAN (Convolutional Neural Network with Generative Adversarial Network), GAN, and Logistic Regression models.

What dataset is used to assess intrusion detection performance?

Performance is evaluated using the CIC IoT Dataset 2023.

Why are SIoT intrusion detection challenges more difficult than traditional settings?

Device heterogeneity, vast traffic generation, decentralized dynamics, and the social/human interaction context increase the likelihood of hidden malicious activity and complicate distinguishing legitimate behavior from threats.

Aug 3, 2026

Upload by Ava Thompson