FedClean: A Decentralized Defense Mechanism Against Parameter Poisoning Attacks in Federated Learning System

  • Position: Machine Learning Research Engineer
  • Type: Part-Time
  • Duration: August 2019 - May 2020
  • Supervisor: Dr. Pan Hui
  • Laboratory: Systems & Media Lab (SymLab)
  • Location: Hong Kong University of Science and Technology, Kowloon, Hong Kong
  • Visit: http://symlab.ust.hk/

Worked as a sole engineer in collaboration with multiple professors (Dr. Pan Hui, Dr. Dimitris Chatzopoulos) and a Ph.D. student. Individually designed and implemented the entire pipeline including the federated learning system (PyTorch), synthetic poisoning attacks, agent sampling mechanism, defense, and multiple smart contracts (Solidity). Developed multiple API’s, extensive documentation, and manually debugged various experimental libraries to produce a fully functional system. Co-authored a conference paper submitted to INFOCOM 2020