Dependable Analysis of Network Application Data

On one hand, the variety and complexity of modern Internet traffic exceeds anything previously imagined. On the other hand, all this variety is embedded into web data where different applications and protocols get nested inside one another. This means that differentiating and isolating different behaviours will be very challenging. Realizing the goal of analyzing such network and application data is critical for reliable and secure network / service operations for any organization. In this research, we are specifically interested in modelling and understanding the impact of changes in two critical network and application functionalities, namely Hyper-Text Transfer Protocol (HTTP) and Domain Name System (DNS). HTTP is an application-level protocol for distributed, collaborative, hypermedia information systems. DNS is a critical network functionality in the form of a criticial application that maps human usable domain names to machine usable network addresses (and vice versa). Recently, both of these applications have been going through many changes. They are evolving on a day to day basis with new protocols and services embedded in them. Thus, when you compare them to their earlier versions at the beginning of the Internet, you can see that they are going through a revolution. This potentially implies that significant changes will result in the web and domain name to IP address mapping services. These new versions (variants) will change the connection management layer, as well as changing the nature of the traffic and application behaviours. Furthermore, this threads the entire Internet through the eye of HTTPS into which everything seems to be encapsulated. As a consequence, different application and network behaviours will become more and more opaque. This means that traditional monitoring and analysis approaches for keeping networks and services healthy will no longer be possible. Adding to this mix are network address translation and proxy services, streaming audio / video and interactive peer-to-peer applications, all of which increase the challenges of analyzing the network and application data. This then will challenge our ability to identify behaviours leading to cyber faults and cyber threats for different applications and services. Thus, the proposed research program aims to create the foundations and systems for the successful deployment of intelligent networks in practice. It will advance the state-of-the-art in network and application data analysis via robust and dependable machine learning based techniques. It will produce HQP by addressing the talent gap for the next wave in innovations and entrepreneurship of dependable analysis of network application data. In return, this will directly benefit Canada by reducing network downtime and network expenditures for operating healthy networks and applications.