Reproduce “PrivEdge: From Local to Distributed Private Training and Prediction” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Ali Shahin Shamsabadi, Adria Gascon, Hamed Haddadi, Andrea Cavallaro

PrivEdge lets users train one-class image reconstructors locally and uses two non-colluding providers to compare secret-shared images and model parameters during prediction. It reveals the classification output and assumes the providers do not collude. On IAM handwritten text, it reports 92% recall and 91% precision, each 4% lower than a centralized non-private method; performance is comparable on two other tasks.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.