Abstract:With the proliferation of IoT devices, researchers have developed a variety of IoT device identification methods with the assistance of machine learning. Nevertheless, the security of these identification methods mostly depends on collected training data. In this research, we propose a novel attack strategy named IoTGAN to manipulate an IoT device's traffic such that it can evade machine learning based IoT device identification. In the development of IoTGAN, we have two major technical challenges: (i) How to obtain the discriminative model in a black-box setting, and (ii) How to add perturbations to IoT traffic through the manipulative model, so as to evade the identification while not influencing the functionality of IoT devices. To address these challenges, a neural network based substitute model is used to fit the target model in black-box settings, it works as a discriminative model in IoTGAN. A manipulative model is trained to add adversarial perturbations into the IoT device's traffic to evade the substitute model. Experimental results show that IoTGAN can successfully achieve the attack goals. We also develop efficient countermeasures to protect machine learning based IoT device identification from been undermined by IoTGAN.

What problem does this paper attempt to address?

### Problems Addressed by the Paper This paper aims to explore how Internet of Things (IoT) devices can evade machine learning-based device identification methods by manipulating their network traffic. Specifically, the paper proposes a new attack strategy called IoTGAN, which leverages the concept of Generative Adversarial Networks (GANs) to enable malicious IoT devices to disguise their traffic, thereby avoiding detection by existing machine learning models. #### Main Research Content: 1. **Analysis of Existing Identification Methods**: - The study examines existing machine learning-based IoT device identification methods and points out potential vulnerabilities in black-box environments. 2. **Design and Implementation of IoTGAN**: - A new attack strategy, IoTGAN, is proposed, which includes two main components: the Discriminative Model and the Manipulative Model. - The Discriminative Model is used to simulate the behavior of the target identification model; the Manipulative Model is used to generate adversarial perturbations so that the IoT device's traffic can evade detection. 3. **Technical Challenges**: - How to obtain the Discriminative Model in a black-box environment. - How to add perturbations without affecting the functionality of the IoT device. 4. **Defense Measures**: - A Device Profiling method is proposed, which utilizes inherent hardware characteristics in wireless signals to resist IoTGAN attacks, ensuring the accuracy of device identification. #### Experimental Results: - Experiments demonstrate that IoTGAN can successfully bypass various machine learning-based identification methods, with an attack success rate of over 90%. - After deploying Device Profiling, the attack success rate significantly drops to near zero. ### Summary This paper aims to reveal the security vulnerabilities of existing machine learning-based IoT device identification methods and proposes effective attack and defense strategies, providing important references for future IoT security research.

IoTGAN: GAN Powered Camouflage Against Machine Learning Based IoT Device Identification

Adversarial Example Attacks in Internet of Things (IoT)

Toward Efficiently Evaluating the Robustness of Deep Neural Networks in IoT Systems: A GAN-Based Method

A GAN-Based Defense Framework Against Model Inversion Attacks.

Secure Collaborative Deep Learning Against GAN Attacks in the Internet of Things

Anomaly-Based Intrusion on IoT Networks Using AIGAN-a Generative Adversarial Network

Intelligent Intrusion Detection for Internet of Things Security: A Deep Convolutional Generative Adversarial Network-enabled Approach

GAN-Based Privacy Abuse Attack on Federated Learning in IoT Networks

Hybrid Data Augmentation Based Machine Learning Approach for Botnet Attack Detection in IOT Networks

Toward Generative Adversarial Networks for the Industrial Internet of Things

IoT Network Security from the Perspective of Adversarial Deep Learning

IoT Device Identification Based on Network Communication Analysis Using Deep Learning

Obfuscating IoT Device Scanning Activity via Adversarial Example Generation

TMG-GAN: Generative Adversarial Networks-Based Imbalanced Learning for Network Intrusion Detection

Mitigate noisy data for smart IoT via GAN based machine unlearning

iPET: Privacy Enhancing Traffic Perturbations for Secure IoT Communications

IoT Device Identification Using Deep Learning

Knowledge Enhanced GAN for IoT Traffic Generation

Machine learning based intrusion detection framework for detecting security attacks in internet of things

Enhancing Cybersecurity in IoT Networks: A Deep Learning Approach to Anomaly Detection

Your Smart Home Can't Keep a Secret: Towards Automated Fingerprinting of IoT Traffic with Neural Networks