Dissertation · 2024 · Open access

Recovering Utility in LDP Schemes by Training with Noise^2

Kexin Li

TSpace (University of Toronto) · Published 1 November 2024

Open PDFView at the publisherCite this paper

In brief

“Recovering Utility in LDP Schemes by Training with Noise^2” is a 2024 dissertation by Kexin Li, published in TSpace (University of Toronto). OpenAlex counts 0 citations as of 19 November 2026. It is open access under a CC BY licence: the full text is free to read at TSpace (University of Toronto).

Citations
0
References
95
Authors
1

Abstract

The adoption of large cloud-based models for inference has been hampered by concerns about the privacy leakage of end-user data. One method to mitigate this leakage is to add local differentially private noise to queries before sending them to the cloud, but this degrades utility as a side effect. Our key insight is that knowledge available in the noisy labels returned from performing inference on noisy inputs can be aggregated and used to recover the correct labels. We implement this insight in LDPKiT, which stands for Local Differentially-Private and Utility-Preserving Inference via Knowledge Transfer. LDPKiT uses the noisy labels returned from querying a set of noised inputs to train a local model (noiseˆ2 ), which is then used to perform inference on the original set of inputs. Our experiments on CIFAR-10, Fashion-MNIST, SVHN, and CARER NLP datasets demonstrate that LDPKiT can improve utility without compromising privacy. For instance, on CIFAR-10, compared to a standard ϵ-LDP scheme applying the Laplacian noise on each data sample with ϵ = 15, which provides a weak privacy guarantee, LDPKiT can achieve similar accuracy with ϵ = 7, offering an enhanced privacy guarantee. Moreover, the benefits of using LDPKiT increase at higher, more privacy-protective noise levels. For Fashion-MNIST and CARER, LDPKiT’s accuracy on the sensitive dataset with ϵ = 7 not only exceeds the average accuracy of the standard ϵ-LDP scheme with ϵ = 7 by roughly 20% and 9% but also outperforms the standard ϵ-LDP scheme with ϵ = 15, a scenario with less noise and minimal privacy protection. We also perform Zest distance measurements to demonstrate that the type of distillation performed by LDPKiT is different from a model extraction attack.

Cite this paper

Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.

APA 7

Li, K. (2024). Recovering Utility in LDP Schemes by Training with Noise^2. TSpace (University of Toronto). http://hdl.handle.net/1807/140918

MLA 9

Li, Kexin. “Recovering Utility in LDP Schemes by Training with Noise^2.” TSpace (University of Toronto), 2024. http://hdl.handle.net/1807/140918.

Chicago

Li, Kexin. 2024. “Recovering Utility in LDP Schemes by Training with Noise^2.” TSpace (University of Toronto). http://hdl.handle.net/1807/140918.

Harvard

Li, K. (2024) ‘Recovering Utility in LDP Schemes by Training with Noise^2’, TSpace (University of Toronto). Available at: http://hdl.handle.net/1807/140918.

IEEE

K. Li, “Recovering Utility in LDP Schemes by Training with Noise^2,” TSpace (University of Toronto), 2024.

BibTeX

@phdthesis{li2024recovering,
  title     = {{Recovering Utility in LDP Schemes by Training with Noise^2}},
  author    = {Li, Kexin},
  year      = {2024},
  url       = {http://hdl.handle.net/1807/140918}
}

Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)

Topic and keywords

Physical Sciences › Computer Science › Artificial Intelligence › Privacy-Preserving Technologies in Data

  • local differential privacy
  • label noise
  • knowledge distillation
  • privacy-utility trade-off
  • label recovery
  • Laplacian noise
  • model extraction attacks

Work with this paper

Most-cited papers in Computer Science

All highly cited Computer Science papers

Questions about this paper

How do I cite this paper in APA style?

Li, K. (2024). Recovering Utility in LDP Schemes by Training with Noise^2. TSpace (University of Toronto). http://hdl.handle.net/1807/140918

How many times has this paper been cited?

0 times, according to OpenAlex on 19 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.

Is this paper free to read?

Yes. It is open access at TSpace (University of Toronto): sign in and open the PDF from this page.

Who wrote this paper?

Kexin Li.

Bibliographic data from OpenAlex (CC0), last checked 19 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.