News

  • 12/2018: Our recent work “Minimizing Trust in Hardware Wallets with Two Factor Signatures” has been accepted to FC 2019

  • 12/2018: I will be presenting “Outsourcing Private Machine Learning via Lightweight Secure Arithmetic Computation” at the PPML workshop @NeurIPS

  • 04/2017: Our work on Secure Aggregation for Federated Machine Learning was featured in the Google Research Blog

Publications

Minimizing Trust in Hardware Wallets with Two Factor Signatures
FC 2019 (to appear)

Outsourcing Private Machine Learning via Lightweight Secure Arithmetic Computation
PPML 2018 (NeurIPS workshop)

PDF

Practical secure aggregation for privacy-preserving machine learning
CCS 2017

PDF

Bounded KDM security from iO and OWF
SCN 2016

PDF

Secure Dating with Four or Fewer Cards.

PDF

Authenticating computation on groups: New homomorphic primitives and applications
ASIACRYPT 2014

PDF

Obfuscation $\Rightarrow$ (IND-CPA Security $\not\Rightarrow$ Circular Security)
SCN 2014

PDF

Industry Experience

 
 
 
 
 
May 2018 – August 2018
New York City, USA

Software Engineering Intern

Keybase Inc.

Audited the design of the Keybase app, identifying and fixing some bugs/vulnerabilities (both at the implementation and cryptographic design level). Upgraded and extended the encryption functionality of the Keybase client to allow generating Saltpack encrypted messages for Keybase teams (Golang).
 
 
 
 
 
May 2017 – August 2017
Los Angeles, USA

Software Engineering Intern

Snap Inc.

Contributed to the Google KeyTransparency open source project (prevents MiTM attacks by storing users’ public keys in a transparent auditable log) and started developing an Android client app library/app (Golang/Java). Helped design and implement a system for secure distributed machine learning (Java).
 
 
 
 
 
May 2016 – August 2016
New York City, USA

Software Engineering Intern

Google

Designed and implemented a new prototype Java library that will be used to perform federated machine learning in a privacy preserving way. This will allow Google to train machine learning models from the data of many users (i.e. android phones) without learning the inputs of each user. One example application is improving accuracy of keyboard predictions: the goal is to suggest the next word which a user might want to type without learning what each individual user is typing.

Contact