Projects

• Proposed a new family of algorithms for private, trustless sealed-bid auctions on blockchain, combining Dutch auction model with a novel stepwise revelation tree structure.

• Designed a reveal tree that reduces the number of rounds to O(log n), significantly lowering gas costs and execution times while ensuring bid confidentiality through cryptographic commitments and zero-knowledge proofs.

• Developed a novel dynamic programming algorithm for gas superoptimization of smart contracts, addressing the scalability limitations of existing Max-SMT-based approaches on large bytecode blocks.

• Achieved more than double the gas savings compared to syrup 2.0 alone, reducing the gas usage of real-world smart contracts by 11.23%.

• Designed and implemented a blockchain-based e-voting system utilizing blind signatures and the relayer structure, achieving untraceable anonymity, verifiability, together with strong security, privacy, etc.

• Achieved at least 40% greater efficiency in terms of gas costs comparing to existing systems with similar security measures.

• Implemented a tamper-proof protocol for generating truly uniform random numbers on Ethereum by leveraging RSA, Goldwasser-Micali cryptosystem, and a secret sharing scheme.

• The generated numbers are tamper-proof and remain secret until they are finalized.

• Visualized COVID-19 data—including infection, death, and vaccination rates—in U.S. prisons using various graphs. The system summarizes data by geography, state political affiliation, facility type, and prison personnel.

• Awarded the HKUST Best FYP Award and the Huawei FYP award for the year 22/23.

• Collaborated with FYP team members to design and implement a Vacancy Detection System for a smart car park, achieving 91% accuracy with no false detections in real-world testing.

• The system utilizes transfer learning, CNN, and YOLOv5, and achieves a wider detection angle than industry standards by incorporating a fisheye camera and an undistortion algorithm.

• Developed a price predictor for Apple Inc. (AAPL) by training a multilayer perceptron on historical stock data. The system dynamically forecasts the next-hour stock trend, displayed on an interactive webpage built using AWS for seamless user engagement.