Abstract:Enormous risks and hidden dangers of information security exist in the applications of Internet of Things (IoT) technologies. To secure IoT software systems, software engineers have to deploy advanced security software such as Intrusion Detection Systems (IDS) that are able to keep track of how the IoT devices behave within the network and detect any malicious activity that may be occurring. Considering that IoT devices generate large amounts of data, Artificial intelligence (AI) is often regarded as the best method for implementing IDS thanks to AI's high capability in processing large amounts of IoT data. To tackle these security concerns, specifically the ones tied to the privacy of data used in IoT systems, the software implementation of a Federated Learning (FL) method is often used to improve both privacy preservation (PP) and scalability in IoT networks. In this paper, we present a FL IDS that leverages a 1 Dimensional Convolutional Neural Network (CNN) for efficient and accurate intrusion detection in IoT networks. To address the critical issue of PP in FL, we incorporate three techniques: Differential Privacy, Diffie–Hellman Key Exchange, and Homomorphic Encryption. To evaluate the effectiveness of our solution, we conduct experiments on seven publicly available IoT datasets: TON IoT, IoT-23, Bot-IoT, CIC IoT 2023, CIC IoMT 2024, RT-IoT 2022, and EdgeIIoT. Our CNN-based approach achieves outstanding performance with an average accuracy, precision, recall, and F1-score of 97.31%, 95.59%, 92.43%, and 92.69%, respectively, across these datasets. These results demonstrate the effectiveness of our approach in accurately identifying and detecting intrusions in IoT networks. Furthermore, our experiments reveal that implementing all three PP techniques only incurs a minimal increase in computation time, with a 10% overhead compared to our solution without any PP mechanisms. This finding highlights the feasibility and efficiency of our solution in maintaining privacy while achieving high performance. Finally, we show the effectiveness of our solution through a comparison study with other recent IDS trained and tested on the same datasets we use.

A privacy preservation framework for feedforward-designed convolutional neural networks

A Distributed Privacy-Preserving Framework for Deep Learning with Edge-Cloud Computing.

Industrial Scale Privacy Preserving Deep Neural Network

FPCNN: A fast privacy-preserving outsourced convolutional neural network with low-bandwidth

A Privacy-Preserving Scheme For Convolutional Neural Network-Based Applications In Mobile Cloud

Towards Scalable and Privacy-Preserving Deep Neural Network via Algorithmic-Cryptographic Co-design

Differentially Private Convolutional Neural Networks with Adaptive Gradient Descent.

Privacy‐preserving convolutional neural network prediction with low latency and lightweight users

Synergistic Neuromorphic Federated Learning with ANN-SNN Conversion For Privacy Protection

Toward Scalable and Privacy-preserving Deep Neural Network Via Algorithmic-Cryptographic Co-design

Privacy Prediction of Lightweight Convolutional Neural Network

Masked image: Visually protected image dataset privacy-preserving scheme for convolutional neural networks

LF3PFL: A Practical Privacy-Preserving Federated Learning Algorithm Based on Local Federalization Scheme

PrivyNet: A Flexible Framework for Privacy-Preserving Deep Neural Network Training

Pratical Privacy-preserving Convolutional Neural Network Inference Framework with Edge Computing for Health Monitoring

Enhancing Video Surveillance and Behavior Recognition With Deep Learning While Ensuring Privacy Protection

cuSCNN: A Secure and Batch-Processing Framework for Privacy-Preserving Convolutional Neural Network Prediction on GPU

Towards Enhancing Privacy-Preservation of a Federated Learning CNN Intrusion Detection System in IoT: Method and Empirical Study

FALCON: A Fourier Transform Based Approach for Fast and Secure Convolutional Neural Network Predictions

Toward Secure and Private Over-the-Air Federated Learning

Differentially Private Federated Learning with an Adaptive Noise Mechanism