The aim of the ECODE proposal is to develop, implement and validate experimentally a cognitive routing system. This system will use cognitive techniques to provide a routing system able to meet the future Internet challenges. The project gathers networking and machine learning experts to address these challenges efficiently. In architectural phase one, the cognitive routing system is designed and prototyped. In phase two, we consider three representative sets of use cases to evaluate the benefits of machine learning techniques to improve a) the Internet manageability and security, b) the availability of Internet paths, and c) the scalability and quality of the Internet routing system. Concerning the manageability and security use case, the project will develop, implement and experiment new techniques that allow 1) to adaptively sample traffic on core links, 2) to efficiently monitor the path performance by combining passive and active measurements, and 3) to cooperatively detect traffic anomalies leading to performance or QoS decrease, and cooperatively detect intrusions and attacks. Concerning the availability use case, the project will develop, implement and experiment new techniques to allow 1) efficient path ranking based on QoS and availability metrics, 2) routers to efficiently reroute paths to other links in case of failure, and 3) routers to correlate traffic flows to diagnose and predict deviation over time. Concerning the routing scalability and quality use case, the project will develop, implement and experiment new techniques to detect events that are detrimental to the routing system dynamics and to efficiently react to such events. In phase three, we implement and experiment novel semi-supervised, on-line, and distributed machine learning techniques within the cognitive routing system. The experimentation and the validation of the techniques developed in the project will be carried on physical (iLAB) and virtual (e.g., OneLab) experimental facilities.