GNNs and graph analytics for entity resolution and fraud detection. My current research develops algorithms and graph AI methods for entity resolution and financial crime detection. On the theoretical side, I study the expressivity of message-passing neural networks, characterizing the minimal GNN architectures sufficient for entity resolution and fraud detection. On the applied side, I develop algorithms and graph-based machine learning methods for detecting fraud patterns in financial transaction networks characteristic of money laundering. Across both directions, the unifying theme is understanding and exploiting graph structure — through classical algorithms, GNNs, or their combination — to solve detection and analysis problems on real-world networked data. Planned extensions include anomaly detection in computer networks, where the same graph-structural and GNN methods apply to identifying unusual patterns in network traffic, and graph analytics for telecom network optimization. This research is carried out at the Centre for Cybersecurity, Trust and Reliability (CyStar), CSE Department, IIT Madras, funded by a Dun & Bradstreet CSR grant on master data management. I provide research guidance to MTech project students and Project Associates at CyStar on topics including GNN-based subgraph classification for fraud detection, GNN explainability for financial transaction graphs, and synthetic blockchain data generation.
Research Funding
J. Augustine and A. Ganesan, "Secure and Scalable Master Data Management: Integrating Blockchain, Byzantine Fault Tolerance, and Graph-Based Analytics," Dun & Bradstreet CSR Grant, 2025. Funded.
Combinatorial interference models for wireless communication networks. Investigated performance guarantees of distributed algorithms in wireless networks where each node has only local information. Proved tight bounds for distributed scheduling under graph and hypergraph interference models. Published in IEEE/ACM Transactions on Networking, IEEE Transactions on Information Theory, IEEE Transactions on Mobile Computing, Theoretical Computer Science, Wireless Networks, and Applied Mathematics Letters.
Automorphism Groups of Cayley Graphs and Interconnection Networks. Computed the full automorphism groups of Cayley graphs generated by transpositions, resolving open problems in the literature on the symmetry and fault tolerance of interconnection networks. Published in Discrete Mathematics, Journal of Algebraic Combinatorics, Discussiones Mathematicae Graph Theory, Australasian Journal of Combinatorics, Ars Combinatoria, and Discrete Applied Mathematics.
A. Ganesan,
"A Tight Expressivity Hierarchy for GNN-Based Entity Resolution in Master Data Management,"
Technical Report, 48 pages, March 2026. [ arXiv:2603.27154 ]
A. Ganesan,
"The Structure of Hypergraphs Arising in Cellular Mobile Communication Systems,"
IEEE Transactions on Mobile Computing, vol. 25, no. 1, pp. 150-164, January 2025. [ doi, preprint ]
A. Ganesan,
"Performance analysis of distance-1 distributed algorithms for admission control under the 2-hop interference model,"
Theoretical Computer Science, Vol. 947, Article 113718, 16 pages, 20 February 2023. [ doi ]
A. Ganesan,
"On some distributed scheduling algorithms for wireless networks with hypergraph interference models,"
IEEE Transactions on Information Theory, vol. 67, no. 5, pp. 2952-2957, May 2021. [ preprint , doi ]
A. Ganesan,
"Performance guarantees of distributed algorithms for QoS in wireless ad hoc networks,"
IEEE/ACM Transactions on Networking, vol. 28, pp. 182-195, February 2020. [ doi, PDF, preprint ]
A. Ganesan,
"Fault tolerant supergraphs with automorphisms,"
Discrete Applied Mathematics, vol. 254, pp. 274-279, February 2019. [ doi, preprint ]
A. Ganesan,
"Automorphism group of the complete transposition graph,"
Journal of Algebraic Combinatorics, vol. 42, No. 3, pp. 793-801, November 2015. [ preprint, doi ]
A. Ganesan,
"Automorphism groups of Cayley graphs generated by connected transposition sets,"
Discrete Mathematics, vol. 313, no. 21, pp. 2482--2485, November 2013. [ doi, preprint ]