Research Interests

“Challenged” wireless networks where connectivity is intermittent, resources (e.g. battery, bandwidth) are scarce, and applications of interest have non-stringent real time constraints (e.g. data sensing/collecting, social networking, micro-blogging, online flea markets, etc.). Example networks include sensor networks, vehicular networks, pocket-switched networks, etc. In such networks, one needs to exploit node mobility to bridge disconnected parts of the networks. To this end, controlled redundancy (e.g. replication or coding) along with learning algorithms that infer mobility pattern properties and use them to predict future node contacts have been proven quite useful.

Delay Tolerant (or Opportunistic) Networking

Complex Network Analysis has emerged as a method to improve the performance of large networks arising in computer science. Online Social Networks and the need to scalably analyze and search their connectivity graphs has been a major driving force. Aggregating mobility patterns of nodes to social graphs has also been used successfully to study the growing number of available mobility traces. It enables learning the underlying structure governing human mobility, so as to use it to efficiently "navigate" the sparse connectivity graph.

Social Networks / Complex Network Analysis

An important gap exists between mobility models used in theory and simulations, and real-life mobility characteristics revealed by a number of trace-based analyses. Although some initial efforts have been performed towards bridging this gap, what is really needed is a mobility model that is rich enough to better resemble real-life mobility, yet is analytically tractable. I’m interested in analyzing existing mobility traces from different applications as well as collaborating with researchers who collect new traces, in order to identify the building components for such a model.

Mobility Modeling

It is now becoming too expensive to collect and analyze packet level data, especially closer to the core. Instead, flow data have been proposed to classify traffic, detect anomalies, etc. Yet, new applications and the hiding of them on known (e.g. port 80) or random ports makes these tasks extremely challenging. Cross-correlation between different traffic types, subnet aggregation, and analysis of the communication graphs are some of the methods that can be used to detect anomalies, monitor servers, etc. from a vantage point (e.g. gateway routers).

Traffic Analysis and Host Profiling using Netflow Data

I’m generally interested in the application of different mathematical tools and/or cross-disciplinary approaches to solve or better understand difficult networking problems. Some of these tools include:

· Stochastic Modeling

· Transient Analysis of Random Walks

· Fluid Approximations and Mean-field Analysis

· Statistical Learning Theory

· Complex Network Analysis

· Graph Theory

Analytical Tools for Networks

Thrasyvoulos Spyropoulos

ETH Zurich