PADS is a strategy-proof differentially private auction mechanism designed to allow cloud providers to trade resources with consumers in such a way that individual bidding information of the consumers do not get exposed through the auction mechanism. PADS provides differential privacy and approximate truthfulness guarantees while maintaining good performance in terms of revenue earned and allocation efficiency.

What is the challenge?

  • Protect Each User’s Bidding Information through the Auctioning Mechanism.
  • Randomize to achieve Differential Privacy while also ensuring Higher Resource Allocation Utility.

How the System Works?

  • PADS-ADP Scheme:
    • Iterative Exponential Mechanism: in every iteration, the mechanism chooses one winner from the bidders using an Exponential Mechanism until all bids are selected or all the resources get allocated.
    • Approximate Differential Privacy: PADS-ADP can provide $(\epsilon,\delta)$-differential privacy.
    • Truthfulness: PADS-ADP is truthful independent of the strategies used by the bidders.
  • PADS-DP:
    • Grouping Exponential Mechanism: the bids are grouped by the possible price outcomes.
    • Differential Privacy: PADS-DP can provide $\epsilon$-differential privacy.
    • Approximate Truthfulness: PADS-DP is $\epsilon\Delta$-truthful and it is independentof the strategies used by the bidders.


Jinlai Xu (徐锦来)
Jinlai Xu (徐锦来)

My research interests include Big Model Training/Inference/Finetune Framework, Serverless Computing, Distributed Systems, Fog/Edge and Cloud Computing, Stream Processing Optimization and Blockchain-based Techniques