CAREER: Temporal Network Analysis: Models, Algorithms, and Applications

Temporal networks are a powerful representation structure that support understanding and characterizing various complex systems. Face-to-face human contacts, financial transactions, and computer communications can all be viewed as temporal networks where interactions are active only at certain points in time. Analyzing such networks is important for various applications such as maintaining cyber-secure environments in the context of national security. Another example, in the context of a client-server network of interactions, is determining whether there is a set of servers that interact with clients in an unusually coordinated way. In the context of money laundering, the lifeblood of criminal activities and a source of damages to economic competitiveness in the U.S., can we detect the accounts involved in coordinated cryptocurrency laundering while also performing licit transactions? This project devises a new paradigm for analyzing temporal networks effectively and efficiently and trains next-generation of computer scientists from diverse backgrounds by increasing public scientific engagement, performing outreach to marginalized communities, and course development. In particular, the investigator organizes workshops to reach high-school students in the Buffalo area to inform and educate them about the basics of computer science and network science. Outputs, such as an open-source software framework for temporal network analysis and know-how on critical applications such as intrusion detection and anti-money laundering, are designed to advance and contribute to scientific understanding in various disciplines such as cybersecurity, economics, finance, and social network analysis. This project designs and develops motif-based models and algorithms to analyze and process temporal networks. It will devise generic formalizations in a bottom-up approach by first building primitives in the microscale, then analyzing the subgraphs and periodicity in the mesoscale, and lastly extending the techniques for graphs encountered in real-world applications. This project broadens the knowledge with new models and algorithms that can work on temporal networks with fine resolution and a large timespan. To this end, there are two main research thrusts: (1) a framework for temporal motif analysis; and (2) mesoscale structures and graphs in the wild. The investigator performs theoretical and empirical evaluations for all the proposed models and algorithms. In particular, this project considers two real-world applications in collaboration with industry and government research labs; (1) intrusion detection in bipartite cyber logs; and (2) anti-money laundering in financial and cryptocurrency transactions. This project will make contributions to the fields of graph mining and network science. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.