Privacy Preservation in IoT: Machine Learning Approaches

IoT is experiencing fast proliferation with advancement of computation and communication technologies like edge computing and 5G. Along with this, a mass volume of data is collected and transmitted over IoT, and thereby raise privacy concerns. Sensitive information in the collected data are usually improperly treated without privacy-preserving techniques. Furthermore, data from different sources may be linked to reveal more sensitive information. All these issues put privacy protection under great threats.

In this chapter, we present the mainstream research of current privacy preservation in IoTs built upon the literature review we have done in recent years [1‐3].

Internet of Things (IoT), as the extension of cloud computing and a foundation of edge computing, is fast booming due to its abilities to easy the bothersome issues of networks like high latency, congestion, etc. However, the privacy issues attract increasing concerns which result in the “data isolated islands” while dragging down the performances of IoT.

The cyber-physical social system (CPSS), as an extension of Internet of Things (IoT), maps human social interaction from cyberspace to the physical world by data sharing and posting, such as publishing spatial-temporal data, images, videos, etc. The published data contains individual’s sensitive information, and thereby leads to continues attacks on it. Nowadays, most existing research assumes that the privacy protection should be uniform that all parties share a same privacy protection level. This impractical assumption results in data degradation due to over-protection or privacy leakage due to under-protection.

The smart mobile device, as an indispensable component of IoT, has been growing in volume and diversity in recent years. Along with this, cyber-physical social network (CPSN) experiences fast booming, in which users publish their posts or data for sharing. However, since the published data is usually public to all, adversaries can crawl data or launch attacks without much efforts. Existing research usually considers a static adversary where the attack is launched once to steal a type of sensitive information like identity or location. This is not practical in real-world scenarios. To release this assumption, we develop a hybrid privacy-preserving model that protects identity and location privacy at the same time against a dynamic adversary who actively launches attacks. In the proposed model, the privacy protection problem is also considered as a trade-off optimization model that users target on maximizing data utility with high-level privacy protection but adversaries try to achieve a opposite target. To make this happen, we model a multi-stage game built upon Markov Decision Process (GMDP). The user and the adversaries are regarded as two players (also known as parties) in this dynamic zero-sum game. The output of this game should be the optimal actions of users that an adversary cannot breach more privacy no matter how the actions changed. An improved reinforcement learning algorithm based on state-action-reward-state-action (SARSA) is developed, which reduces the cardinality from n to 2. This can significantly improve the convergence efficiency. At last, we conduct experimental simulations on real-world datasets to testify the superior performance on efficiency and feasibility over existing research. This paper is mainly based on our research on privacy protection using reinforcement learning [1‐5].

From Chaps. 2–5, we have shown the existing research status of machine learning driven privacy preservation in IoTs, especially focus on three leading directions using GAN, federated learning, and reinforcement learning. Several advanced technologies and theories are also integrated, for example, differential privacy, game theory, blockchain, etc. Nevertheless, there are still plenty of significant and prospective issues worthy investigating. The popularization of blockchain, digital twin, and artificial intelligence offers a mass of opportunities for researches on machine learning driven privacy preservation in IoTs, but meanwhile they raise new challenges such as unsatisfying data utility and limited communication and computing resources. In addition, there are various other research topics that desiderata consideration in machine learning driven privacy preservation in IoTs. To pave the way for readers and forthcoming researchers, we outline several potentially promising research directions that may be worthy of future efforts.

In this monograph, we summarize the state-of-art research on machine learning driven privacy protection models in IoT scenarios. As far as we can see, machine learning driven privacy protection is still in its early stage. What have been present in this monograph, like models, theories, and potentially conceptual designs, could serve as a point of departure for follow-up readers, students, engineers, and researchers to investigate this emerging field. Our target is to provide a systematic summary of existing research and application outputs on machine learning driven privacy protection in IoTs. Based on this, we analyze the theoretical and practical applicability under big data settings. Subsequently, we offer the interested readers several future research directions, which we hope the following explorers find them insightful or inspiring from a certain angle.