Abstract-The Internet of Things (IoT) is now destroying the barriers between the real and digital worlds. However, one of the huge problems that can slow down the development of this global wave, or even stop it, concerns security and privacy requirements. The criticality of these latter comes especially from the fact that the smart objects may contain very intimate information or even may be responsible for protecting people's lives. In this paper, the focus is on access control in the IoT context by proposing a dynamic and fully distributed security policy. Our proposal will be based, on one hand, on the concept of the blockchain to ensure the distributed aspect strongly recommended in the IoT; and on the other hand on machine learning algorithms, particularly on reinforcement learning category, in order to provide a dynamic, optimized and selfadjusted security policy.

The Internet of Things (IoT) has become a global sensory network that links physical and virtual objects by communicating and exploiting data and initiating physical actions. The evolution of this paradigm is already threatened by security issues, which constitute major risk factors that demand efficient solutions adapted to the IoT context. In this paper, we put forward a logical approach and systemic analysis that enables us to present the key aspects of new access control (AC) model for the IoT environments, called a pervasive-based access control model (PerBAC). Our approach is based on the study of important, reputable AC models that we use as a background for our proposed model. PerBAC is defined here based on a representation of the decision-making algorithm, a description of the abstract entities using the attributes as a fundamental concept and the collaboration aspects necessary to handle the case of multiple organizations. These attributes are the perfect recipient of the information collected by IoT environments from the physical world and allow optimal access control decisions to be taken according to dynamic rules and entities based on the algorithm. Our interpretation of the attributes, the dynamic entities and their exploitation by our proposed algorithm produce a new AC model adapted to the IoT paradigm.

The Internet of things is no longer a concept; it is a reality already changing our lives. It aims to interconnect almost all daily used devices to help them exchange contextualized data in order to offer services adequately. Based on the existing Internet, IoT suffers indisputably from security issues that could threaten its evolution and its users' interests. Starting from this fact, we try to define the main security threats for the IoT perimeter and propose some pertinent solutions. To do so, we first establish a state of the art concerning the IoT definition, protocols, environment, architecture and security. Then, we expose a case study of a standard IoT platform to illustrate the impact of security on all IoT layers. Furthermore, the paper presents the results of a security audit on our implemented platform. Finally, based on our evaluation, we highlight many solutions as well as possible directions for future research.

No abstract

The main challenge facing the Internet of Things (IoT) in general, and IoT security in particular, is that humans have never handled such a huge amount of nodes and quantity of data. Fortunately, it turns out that Machine Learning (ML) systems are very effective in the presence of these two elements. However, can IoT devices support ML techniques? In this paper, we investigated this issue and proposed a twofold contribution: a thorough study of the IoT paradigm and its intersections with ML from a security perspective; then, we actually proposed a holistic ML-based framework for access control, which is the defense head of recent IT systems. In addition to learning techniques, this second pillar was based on the organization and attribute concepts to avoid role explosion problems and applied to a smart city case study to prove its effectiveness.