The Internet of Things, its application and their QoS requirements. The future Internet will include not only usual terminals but more generally any form of connected objects (or things) authorizing the development of new business activities, in various domains such as remote supervision, personal assistance, or urban transport. This IoT will also have to meet non-functional needs (e.g. quality of service – QoS, security) of these new applications. The interactions between the underlying application software(s) and the connected objects will be based on heterogeneous networks and on middleware layers. Indeed, from 2010, a major standardization effort has been conducted, notably via the ETSI and the oneM2M consortium1, 2 . The resulting frameworks are aimed at abstracting applications from complexity of the underlying technologies (networks and objects); they are also aimed at avoiding vertical fragmentation of currently developed IoT solutions thanks to a generic middleware layer. Based on the REST architectural style (http, CoAP, …), this framework makes them appear what can be called Middleware (MW) nodes, named gateway and server in the ETSI vision, and mn-cse and in-cse in the oneM2M vision (Fig. 1). Both visions make also appear two major bottlenecks with regard to QoS considerations (service availability, bounded response time, etc.): within the connected objects and IP networks, and within the MW nodes. In this context, several attempts have already been done to face (among other) the QoS requirements at the middleware level. Those attempts are based, for instance, on the deployment of QoS-oriented mechanisms within the MW nodes, with the aim to manage the application traffic (e.g. delaying less priority http requests in case of congestion) and / or allocated resources of the underlying machines3 . Those propositions have also shown the benefits that can be induced by a deployment of such mechanisms “outside” of the data path, for instance thanks to intermediate proxies configured (for instance) as traffic load balancer / shaper / dropper (Fig. 2).

Many vulnerabilities affect applications based on the Internet of Things (IoT) paradigm. This is mainly due to the wireless nature of the communications taking place among the involved entities, which can be represented by heterogeneous devices, such as sensors, actuators, RFID, NFC, and so on. As a consequence, an IoT-based system can be object of various kinds of attacks to the network resources, to the integrity and confidentiality of the transmitted data, or to the privacy of the users or companies that exploit IoT services. Note that security is one of the main obstacles towards the wide adoption and diffusion of IoT applications [1]. Until now, little attention has been paid towards the identification and counteraction of specific kinds of attacks pursued in IoT contexts, but several solutions are available for wireless sensor networks (WSNs) [2] and mobile ad hoc networks (MANETs) [3], as presented in the state of the art section. However, such approaches should be revised with respect to the dynamicity and heterogeneity of the IoT environment. In order to guarantee proper levels of security, an IoT system should provide authentication systems for: (i) the entities that send data to the IoT platform; (ii) the entities that require information to the IoT platform in the form of services; then it should also provide encryption mechanisms for ciphering the transmitted data and, to do this, adopt an effective key management system. Finally, policies able to regulate the access to the available resources should be defined and enforced. However, besides such countermeasures would improve the resilience of the IoT system, particular attention must be directed against Denial of Service (DoS) attack, which is widely considered to be an important security issue [4]. There are various kinds of DoS attack, such as flooding the network with useless traffic, resource power exhaustion, and jamming the signal [5]. In fact, they aim at making devices or network resources unavailable by preventing the communications and/or compromising the traffic or wasting resources and, thus, blocking the access to the services provided by the IoT platform. Note that generally no variation is made on the transmitted data by such a kind of attack. DoS attacks can be initialized from a remote place by means of proper tools and commands. In the IoT, the situation worsens if not only data sources are attacked but if also the IoT platform itself is picked on. If the IoT platform, that acts as a distributed server, is saturated, for example, with external requests, then it cannot respond to legitimate ones in a reasonable time (i.e., because communications are slowed by the attack). On the other side, storage and computational resources of both devices and IoT platform will be consumed, thus generating further damages [6].

Seventh International Conference on Measuring Technology and Mechatronics Automation