Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

OptiFLIDS: Optimized Federated Learning for Energy-Efficient Intrusion Detection in IoT

Saida Elouardi, Mohammed Jouhari, Anas Motii

TL;DR

OptiFLIDS tackles intrusion detection in IoT under privacy and resource constraints by integrating pruning directly into Federated Learning and guiding pruning decisions with a Deep Reinforcement Learning agent. It models non-IID client data using a Gamma-based partitioning scheme, prunes local CNNs with an early-one-shot mask, and aggregates only unpruned weights via a specialized server-side rule, balancing accuracy and energy through a multi-objective objective. Evaluations on Ton_IoT, X-IIoTID, and IDSIoT2024 show that substantial pruning (often >60%) can be achieved with minimal accuracy degradation, and DRL optimization yields notable energy savings while maintaining robust detection across IID and non-IID settings. The work demonstrates practical viability for deploying energy-efficient, privacy-preserving IDS in real-world, resource-limited IoT environments.

Abstract

In critical IoT environments, such as smart homes and industrial systems, effective Intrusion Detection Systems (IDS) are essential for ensuring security. However, developing robust IDS solutions remains a significant challenge. Traditional machine learning-based IDS models typically require large datasets, but data sharing is often limited due to privacy and security concerns. Federated Learning (FL) presents a promising alternative by enabling collaborative model training without sharing raw data. Despite its advantages, FL still faces key challenges, such as data heterogeneity (non-IID data) and high energy and computation costs, particularly for resource constrained IoT devices. To address these issues, this paper proposes OptiFLIDS, a novel approach that applies pruning techniques during local training to reduce model complexity and energy consumption. It also incorporates a customized aggregation method to better handle pruned models that differ due to non-IID data distributions. Experiments conducted on three recent IoT IDS datasets, TON_IoT, X-IIoTID, and IDSIoT2024, demonstrate that OptiFLIDS maintains strong detection performance while improving energy efficiency, making it well-suited for deployment in real-world IoT environments.

OptiFLIDS: Optimized Federated Learning for Energy-Efficient Intrusion Detection in IoT

TL;DR

OptiFLIDS tackles intrusion detection in IoT under privacy and resource constraints by integrating pruning directly into Federated Learning and guiding pruning decisions with a Deep Reinforcement Learning agent. It models non-IID client data using a Gamma-based partitioning scheme, prunes local CNNs with an early-one-shot mask, and aggregates only unpruned weights via a specialized server-side rule, balancing accuracy and energy through a multi-objective objective. Evaluations on Ton_IoT, X-IIoTID, and IDSIoT2024 show that substantial pruning (often >60%) can be achieved with minimal accuracy degradation, and DRL optimization yields notable energy savings while maintaining robust detection across IID and non-IID settings. The work demonstrates practical viability for deploying energy-efficient, privacy-preserving IDS in real-world, resource-limited IoT environments.

Abstract

In critical IoT environments, such as smart homes and industrial systems, effective Intrusion Detection Systems (IDS) are essential for ensuring security. However, developing robust IDS solutions remains a significant challenge. Traditional machine learning-based IDS models typically require large datasets, but data sharing is often limited due to privacy and security concerns. Federated Learning (FL) presents a promising alternative by enabling collaborative model training without sharing raw data. Despite its advantages, FL still faces key challenges, such as data heterogeneity (non-IID data) and high energy and computation costs, particularly for resource constrained IoT devices. To address these issues, this paper proposes OptiFLIDS, a novel approach that applies pruning techniques during local training to reduce model complexity and energy consumption. It also incorporates a customized aggregation method to better handle pruned models that differ due to non-IID data distributions. Experiments conducted on three recent IoT IDS datasets, TON_IoT, X-IIoTID, and IDSIoT2024, demonstrate that OptiFLIDS maintains strong detection performance while improving energy efficiency, making it well-suited for deployment in real-world IoT environments.

Paper Structure

This paper contains 16 sections, 21 equations, 13 figures, 3 tables.

Table of Contents

  1. Introduction
  2. RELATED WORK
  3. Federated Learning Approaches for Intrusion Detection
  4. Compression-Enhanced Federated Learning for IDS
  5. PROPOSED SYSTEM MODEL
  6. Modeling Non-IID Data Using Gamma-Based Partitioning
  7. Model Pruning
  8. The Proposed FL Framework
  9. Model Complexity Analysis
  10. Problem Formulation
  11. Using Deep Reinforcement Learning For Optimization
  12. EXPERIMENTAL SETUP AND PERFORMANCE EVALUATION
  13. Selected Datasets
  14. Experimental Settings
  15. Results Evaluation And Discussion
  16. ...and 1 more sections

Figures (13)

  • Figure 1: Overview of federated intrusion detection in Resource-constrained IoT environments
  • Figure 2: Heatmap of data distribution across clients in the Ton_IoT Dataset Simulated Using Normalized Gamma Distribution with Varying Shape Parameters $\boldsymbol{\alpha}$.
  • Figure 3: Overview of the proposed approach for our federated pruning during the first round of federated learning
  • Figure 6: Proposed model architecture for IDS
  • Figure 7: Impact of pruning on model accuracy across three IoT datasets under IID settings using FedAvg
  • ...and 8 more figures