Research Interest
My specialisation is on cryptography, a key enabling technology for cybersecurity. In cryptography, my primary focus is on secure multi-party computation (MPC), the standard bearer and holy-grail problem, that permits a collection of data-owners to compute a collaborative result, without any of them gaining any knowledge about the data provided by the other, except what is derivable from the final result of the computation. MPC finds application in any scenario that involve computations on sensitive data from two or more entities. Till date, it has shown demonstrable success in several real-life scenarios, with significant payoff to society. For instance, it has been used-- (a) to securely analyze the sensitive salary data of more than 10 millions of employees in the Greater Boston Area in order to calculate pay disparity across gender and race; (b) to train a model on private medical data held by several sources to offer best treatment for diseases like HIV, skin cancer, retinopathy; (c) to compute the probability of two satellites colliding in the space for satellites owned by competing countries; (d) to implement secure auction to find a fair price for sugar-beet in Denmark; (e) to implement online sexual assault reporting platform (allegation escrow) that will detect repeat perpetrators and create pathways to support for victims. Other compelling uses of MPC include disease surveillance, electricity trading markets, scientific discovery, smart-cities, genomics, homeland and cyber security, global advanced persistent threat identification in corporate network data, tax fraud detection and the numerous applications in medicine, finance sector, self-driven automobiles that fall under secure machine learning and prediction.
My secondary interest lies in the area of fault-tolerant distributed computing that includes classic problems such as Byzantine Agreement aka BA (and its lose relative broadcast) that allows a set of distrusting parties to jointly reach agreement on their private inputs even in the face of a coalition of cheating parties. BA has been used to build {\em robust} systems since long. Its solutions have been leading their power in systems ranging from flight control, to databases, to peer-to-peer; Microsoft uses BA in Farsite; many structured peer-to-peer systems use BA. Both broadcast and BA also serve as important building block of MPC. Lastly and importantly, BA has reappeared in a new avatar in the form of Block-chain technology.
The core focus of my research can thus be broadly classified into two areas as follows: (a) Theory and Practice of MPC; (b) Fault-tolerant Distributed Computing. The main goal and the publications under each category is given below.
- Theory and Practice of MPC: The foundational questions for MPC and its building blocks such as circuit garbling, oblivious transfer (OT), commitment schemes, zero-knowledge protocols, verifiable secret sharing (VSS), public key encryptions (PKE), are concerned with the feasibility of realizing these tasks, finding inherent lower bounds on the resources needed for solving these tasks and finding resource-efficient constructions. The resource required by a cryptographic protocol is determined by its computation, round and communication complexity. Building practically-efficient constructs for MPC, their proof-of-concept implementations, performance analysis on specific tasks (such as training a ML model for breast cancer/retinopathy/skin cancer securely, performing secure prediction, e-voting) is the primary concern in the regime of MPC in Practice.
- Fault-tolerant Distributed Computing: This area includes classic problems such as Byzantine Agreement and message communication over untrusted network. Apart from BA and broadcast, I also have worked on the problem of reliable message transfer (RMT) over untrusted network in this domain. My focus involves foundational feasibility, efficiency and optimality questions, in terms of resources such as round (running time), communication and computation, as in the MPC domain.