Anwitaman Datta

My research asks how complex digital systems are structured, how they fail, and how they can be made resilient, trustworthy and accountable. I work at the intersection of distributed systems, graph and network analysis, security and privacy, and the societal implications of digital technologies, combining algorithmic and systems rigour with empirical study of real, messy ecosystems. I come from a background in large-scale resilient distributed systems. Several of my contributions have opened up follow-on lines of research: decentralized online social networks for privacy and censorship resistance; self-repairing (locally repairable) erasure codes and other codes tailored to distributed storage, including elastic redundancy, pipelined archival, and compact storage of multiple data versions by exploiting sparsity; quorum systems for consistency and concurrency control over coded data; and the use of machine learning and stereotypes for computational trust. This work has been recognised with four best paper awards, has led to patents, and was the basis of a storage start-up I co-founded. Along the way I have also worked on complex-network analysis (including entropy-based centrality and clustering for flow networks, applied to cryptocurrency forensics), social media analysis, privacy-preserving data management, cyber-risk insurance, and decision-support systems.

My current research increasingly focuses on the resilience of interconnected digital ecosystems, in which services, platforms, data and AI components depend on one another in ways that are indirect, conditional, shared and only partly documented. One direction asks what we need to know about dependencies to assess resilience reliably: which uncertainties in the dependency structure actually change a conclusion, and how to acquire the evidence that settles them. A second direction treats AI ecosystems as a new kind of networked infrastructure. Models are derived from other models, trained on shared data, and composed into pipelines and agent systems, so failures can be inherited, silently shared across seemingly independent components, and costly. Here I study how defects travel through derivation and coupling, how much redundancy survives shared heritage, and how to quantify and reduce the resulting systemic risk. Across these directions I draw on graph algorithms, information-theoretic and structural analysis, reliability and risk modelling, and empirical experimentation, with the aim of replacing assumptions about how digital systems depend on one another with evidence. The work connects the foundations of distributed systems with the engineering, security and governance of digital infrastructure, including in regulated and government settings.

Most of my publications are listed at: DBLP | Google Scholar | ORCID
Selected talk (slides): Building Trustworthy Decentralized Infrastructure | Cybersecurity: Post Breach | On Ransomware, Cyber-armies and Malicious AI | Cryptocurrency Network Analysis & Forensics | Blockchain for Government