A5077898090- Cryptography and Data Security
- Blockchain Technology Applications and Security
- Distributed systems and fault tolerance
- Adversarial Robustness in Machine Learning
- Advanced Neural Network Applications
University of Maryland, College Park
2022-2024
A zero-knowledge proof of training (zkPoT) enables a party to prove that they have correctly trained committed model based on dataset without revealing any additional information about the or dataset. An ideal zkPoT should offer provable security and privacy guarantees, succinct size verifier runtime, practical prover efficiency. In this work, we present Kaizen, targeted for deep neural networks (DNNs) achieves all these goals at once. Our construction iteratively train their via...
Most blockchain-based cryptocurrencies suffer from a heavily limited transaction throughput, which is barrier to their growing adoption. Payment channel networks (PCNs) are one of the promising solutions this problem. PCNs reduce on-chain load transactions and increase throughput by processing many payments off-chain. In fact, any two users connected via path payment channels (i.e., joint addresses between end-points) can perform payments, underlying blockchain used only when there dispute...