Technologies used: Tensorflow (Keras), LSTM, LSTM-Split models, differential privacy, privacy-preserving noise mechanisms (Gaussian, etc.)
Enabling Trust in ML Models on Untrusted Edges
Worked alongside PhD mentor in Winter 2022-2023. Proposed a long short-term split (LSTM-SPLIT) machine learning model to help obfuscate private data sent between client-server models on untrustworthy edge nodes.