N G supervisor Prof . Mauro Conti University of Padua candidate co - supervisor

The current Internet architecture was designed as a mean of connecting pairs of hosts and allow them to reliably exchange packets. The way we use it every day deeply changed from pure communication to (mostly) content distribution. The wish to fill the gap between the underlying architecture and the information-centric nature of the current Internet traffic has inspired many new architectures with the goal of replacing the current infrastructure. Content-Centric Networking (CCN) is one of them. In CCN, named content becomes a first-class citizen. In contrast with IP, addresses are not explicit. Content consumers simply request what they want to retrieve by name; the network is in charge of delivering the correct content. The decoupling of content and its location allows, among other things, the implementation of ubiquitous caching. Named Data Networking (NDN) is a prominent implementation of the CCN paradigm. NDN’s reliance on caches for performance, reliability and data replication allows an adversary to perform attacks on the architecture that are very effective and relatively easy to implement. This thesis focuses on cache pollution attacks. In these attacks, the adversary’s goal is to alter cache locality, increasing cache misses and link utilization for honest consumers. We review and evaluate previous work on these attacks and their countermeasures. To perform a meaningful analysis, we simulate Internet traffic with appropriate models on a realistic network: the German Research Network (DFN). Finally, given their importance on cache resilience, we perform pollution attacks on caches that rely on different replacement policies. Our results show that such attacks can be implemented in NDN using limited resources, and that their effectiveness is not limited to small topologies. We illustrate that countermeasures designed for IP are not easily applicable to NDN. On the other hand, existing mitigation techniques targeted to NDN provide only limited relief against realistic adversaries. We propose and evaluate a new technique, designed to detect high and low rate cache pollution attacks. Our technique is based on the analysis of variation in traffic distribution as seen by individual routers. We discuss the feasibility of this technique and suggest possible improvements. Part of this work has been submitted to the Elsevier Computer Networks Journal (Special Issue on Information Centric Networking) [19].