Function Secret Sharing for Mixed-Mode and Fixed-Point Secure Computation

Elette Boyle, Nishanth Chandran, Niv Gilboa, Divya Gupta, Yuval Ishai, Nishant Kumar & Mayank Rathee
Published at: Eurocrypt, 2021

We present new and improved FSS gate constructions for applications motivated by ML. More specifically, we improve upon the best previous key sizes for Distributed Comparison Function and also provide new FSS gates for functions which are useful in mixed-mode/fixed-point computation. Read more

[eprint] [bib]

CrypTFlow2: Practical 2-Party Secure Inference

Deevashwer Rathee, Mayank Rathee, Nishant Kumar, Nishanth Chandran, Divya Gupta, Aseem Rastogi & Rahul Sharma
Published at: ACM Conference on Computer and Communications Security (CCS), 2020

We present CrypTFlow2, a cryptographic framework for secure inference over realistic Deep Neural Networks (DNNs) using secure 2-party computation. Read more

[eprint] [bib] [code]

A Practical Model for Collaborative Databases: Securely Mixing, Searching and Computing

Shweta Agrawal, Rachit Garg, Nishant Kumar & Manoj Prabhakaran
Published at: European Symposium on Research in Computer Security (ESORICS), 2020

In this work we introduce the notion of Functionally Encrypted Datastores, which allows multiple data owners to pool in data anonymously at two non-colluding servers and later allows malicious clients to perform search-and-compute queries on the collected data. Read more

[eprint] [bib] [code]

CrypTFlow: Secure TensorFlow Inference

Nishant Kumar, Mayank Rathee, Nishanth Chandran, Divya Gupta, Aseem Rastogi & Rahul Sharma
Published at: IEEE Symposium on Security and Privacy (S&P), 2020

We present CrypTFlow, a first of its kind system that converts TensorFlow inference code into Secure Multi-party Computation (MPC) protocols at the push of a button. The system enables secure inference on real-world networks like ResNet50 over the ImageNet dataset with running time of about 30 seconds for semi-honest security and under 2 minutes for malicious security. Read more

[eprint] [bib] [code]