Nano Autonomous Systems for a totally Decentralized Internet – nano-NET

In the last years has emerged the so-called Internet of Things (IoT) which consists in inter-connecting new (Smart) Objects via the hierarchical AS-graph of the Internet. Complex Human Cyber Physical Systems (HCPS) now require sophisticated interactions between sensors, actuators and humans to take smart decisions while, today, sensors simply push their data to a cloud. More and more computing and storage resources are available in the neighborhood of an object, while pushing the processing in the cloud degrades the performance of real-time applications. Finally, clouds exhibit severe threats on privacy.

In nano-NET, we aim at placing the data as the key component of a new dedicated flat and spontaneous infrastructure inter-connecting Objects. This infrastructure will be made of Objects, which handle and transform the data before forwarding them. Moreover, as more than 20 billions objects are expected to be connected by 2020, private data are increasingly subject to theft, as observed by many great companies. nano-NET will help to design an architecture dedicated to these devices, simplifying their management while guaranteeing their privacy.

We propose here to create a network formed of what we call nano Autonomous Systems (nano- AS) : each nano-AS represents the home of a piece of data (e.g., a measurement, a video, a document). A nano-AS includes a set of devices that relay, process and use the data that has been generated or stored within their area. A nano-AS collects the objects which are part of the same community (e.g. ownership (financial), geographic location, social relationship (a family), membership (a collaborative project)). All of these devices also share a certain level of trust across a nano-AS: they together define how to securely exchange data and what rules need to be enforced to maintain privacy.

The underlying topology has a strong impact on the routing performance (e.g. P2P). Thus, classical blind routing solutions are inaccurate. First, we will propose mechanisms to handle dynamic topologies (e.g. mobile objects, connected directly or through a transit nano-AS). We have also to spatially balance the load: since we will rely on multiple technologies, which potentially mutually interfere if they are wireless, we have to spread the traffic to share efficiently the radio spectrum. We must respect in parallel the QoS requirements (delay, reliability). Finally, the load should be temporally balanced (night/day alternance). For instance, a non critical flow should be able to exploit the silences of other priority flows to save bandwidth and energy.

We want to set-up a collection of consistent routing policies for exchanging private data between nano-AS. Because of the limited capabilities of border routers, they may not be all able to filter the data accurately. Thus, the filtering operations have to be distributed, located close to the source of traffic. Most objects are likely to host several independent applications, owning to different nano-AS. Using routing virtualization mechanisms will help us to deal with such challenges. Because of dense topologies, we have to select cooperatively the best border objects. We have also to provide self-optimizing algorithms: a nano-AS must select the objects which are part of the routing process.

Providing a network topology is not enough to preserve the privacy. We have to incorporate the way to process data (with obfuscation or anonymization). First, we will integrate in the routing policy the way to filter private data. Mechanisms have to be grated to the routing tables to select a correct route, export its accuracy (a nano-AS will select the highest accuracy depending on its relationship/trust). In parallel, we will define a set of primitives to transform the data, incorporated in the routing scheme. We have to be sufficiently generic to be able to handle complex transformation (e.g. video compression / modification, annotation).