Cryptography is more than just hiding information (i.e., encryption). Using cryptography, it is possible to perform a wide range of computations on secret data that cannot be seen. For example, two hospitals may wish to train various models (e.g. classification) on their joint medical database without revealing any information except the output.

The idea of computing on private data is called secure computation which enables distrustful users to jointly evaluate any function on their private inputs without requiring a trusted third party and without revealing anything except the result itself. Trieu is interested in both theoretical and practical aspects of secure computation techniques.  More concretely, her recent research focuses on:

  • Private matching (PM): PM allows several parties to jointly compute certain functions, depending on their private inputs matched in some way. Very recently, PM has been applied to design decentralized contact tracing systems that allow to detect if a person is in the proximity of a COVID19 diagnosed patient, without revealing their location. 

  • Privacy-preserving machine Learning: Many machine learning algorithms require large-scale data. When the data is sensitive and comes from different sources, it is highly desirable to allow different entities to train various models on their joint data while maintaining the privacy of each database.

  • Cryptographic primitives: In addition to concrete optimizations tailored for the real-world and important applications above, Trieu is also interested in making core cryptographic techniques (e.g. oblivious transfer, block cipher) more efficient and quantum-resistant.