Whats Going On? Learning Communication Rules in Edge Networks

Existing traffic analysis tools focus on traffic volume. They identify the heavy-hitters—flows that exchange high volumes of data, yet fail to identify the structure implicit in network traffic—do certain flows happen before, after or along with each other repeatedly over time? Since most traffic is generated by applications~(web browsing, email, p2p), network traffic tends to be governed by a set of underlying rules. Malicious traffic such as network-wide scans for vulnerable hosts~(mySQLbot) also presents distinct patterns. We present eXpose, a technique to learn the underlying rules that govern communication over a network. From packet timing information, eXpose learns rules for network communication that may be spread across multiple hosts, protocols or applications. Our key contribution is a novel statistical rule mining technique to extract significant communication patterns in a packet trace without explicitly being told what to look for. Going beyond rules involving flow pairs, eXpose introduces templates to systematically abstract away parts of flows thereby capturing rules that are otherwise unidentifiable. Deployments within our lab and within a large enterprise show that eXpose discovers rules that help with network monitoring, diagnosis, and intrusion detection with few false positives.