@DiffPriv

Website for the differential privacy research community

 Private Island
differentialprivacy.org
Joined July 2020
  • Tweets62
  • Following0
  • Followers1.6K
One-Sided Differential Privacy differentialprivacy.org/one-…

One-Sided Differential Privacy

Differential privacy is defined in terms of pairs of neighboring datasets. That is, \(M\) is \((\varepsilon,\delta)\)-differentially private if, for all measurable events \(T\) and all neighboring...

differentialprivacy.org
1
6
391
Call for Papers - TPDP 2026 differentialprivacy.org/tpdp…

Call for Papers - TPDP 2026

The 12th Workshop on the Theory and Practice of Differential Privacy (TPDP 2026) will take place on June 1 and 2 in Boston, MA. The deadline to submit a 4-page abstract is February 18, 2026 AoE, with...

differentialprivacy.org
2
387
Open Problem: Selection via Low-Sensitivity Queries differentialprivacy.org/open…

Open Problem: Selection via Low-Sensitivity Queries

Two of the basic tools for building differentially private algorithms are noise addition for answering low-sensitivity queries and the exponential mechanism for selection. Could we do away with the...

differentialprivacy.org
1
11
648
Limits of Privacy Amplification by Subsampling differentialprivacy.org/subs…

Limits of Privacy Amplification by Subsampling

In our previous post we gave a brief introduction to privacy amplification by subsampling. The high-level story is that we can make differentially private algorithms faster by runninng them on a...

differentialprivacy.org
13
583
Privacy Amplification by Subsampling differentialprivacy.org/subs…

Privacy Amplification by Subsampling

Privacy Amplification by Subsampling is an important property of differential privacy. It is key to making many algorithms efficient – particularly in machine learning applications. Thus a lot of...

differentialprivacy.org
16
652
Differentially Private Algorithms that Never Fail differentialprivacy.org/fail…

Differentially Private Algorithms that Never Fail

Most differentially private algorithms fail with some nonzero probability. For example, when adding Gaussian or Laplace noise, there is some chance that the noise deviates significantly from its...

differentialprivacy.org
14
793
Tight RDP & zCDP Bounds from Pure DP differentialprivacy.org/pdp-…

Tight RDP & zCDP Bounds from Pure DP

There are multiple ways to quantify differential privacy, including pure DP [DMNS06], approximate DP [DKMMN06], Concentrated DP [DR16,BS16], Rényi DP [M17], Gaussian DP [DRS19], & function-DP...

differentialprivacy.org
3
18
9,700
Differential Privacy retweeted
Gautam Kamath@thegautamkamath2 Jan 2024
Happy new year! As promised, I wrote a blog post on the #NeurIPS2023 Outstanding Paper on privacy auditing, by @shortstein, @srxzr, Matthew Jagielski. Post is on @DiffPriv, 5 min read, written for a general ML audience. No knowlege of privacy needed! Link in next tweet👇
5
16
159
19,530
NeurIPS 2023 Outstanding Paper: Privacy auditing in just one run differentialprivacy.org/neur…

NeurIPS 2023 Outstanding Paper: Privacy auditing in just one run

NeurIPS 2023 just wrapped up, and one of the two outstanding paper awards went to Privacy Auditing with One (1) Training Run, by Thomas Steinke, Milad Nasr, and Matthew Jagielski. The main result of...

differentialprivacy.org
7
37
5,278
Open problem(s) - How generic can composition results be? differentialprivacy.org/open…

Open problem(s) - How generic can composition results be?

The composition theorem is a cornerstone of differential privacy literature. In its most basic formulation, it states that if two mechanisms \(\mathcal{M}_1\) and \(\mathcal{M}_2\) are respectively...

differentialprivacy.org
1
2
20
4,393
Beyond Local Sensitivity via Down Sensitivity differentialprivacy.org/down…

Beyond Local Sensitivity via Down Sensitivity

In our previous post, we discussed local sensitivity and how we can get accuracy guarantees that scale with local sensitivity, which can be much better than the global sensitivity guarantees attained...

differentialprivacy.org
2
26
10,930
Differential Privacy retweeted
OpenDP@opendp_org12 Sep 2023
OpenDP is excited to bring you Library 0.8! Featuring improvements to OpenDP APIs, an early preview of functionality, and quality-of-life improvements. We'll discuss the new release at the OpenDP Community Meeting on September 29! github.com/opendp/opendp

GitHub - opendp/opendp: The core library of differential privacy algorithms powering the OpenDP...

The core library of differential privacy algorithms powering the OpenDP Project. - opendp/opendp

github.com
4
23
2,157
Beyond Global Sensitivity via Inverse Sensitivity differentialprivacy.org/inve…

Beyond Global Sensitivity via Inverse Sensitivity

The most well-known and widely-used method for achieving differential privacy is to compute the true function value \(f(x)\) and then add Laplace or Gaussian noise scaled to the global sensitivity of...

differentialprivacy.org
8
1
26
7,796
Covariance-Aware Private Mean Estimation, Efficiently differentialprivacy.org/colt…

Covariance-Aware Private Mean Estimation, Efficiently

Last week, the Mark Fulk award for best student paper at COLT 2023 was awarded to the following two papers on private mean estimation: A Fast Algorithm for Adaptive Private Mean Estimation, by John...

differentialprivacy.org
9
1
26
12,409
#COLT2023 Mark Fulk award for best student paper is shared by two papers
1
3
2
26
36,259
Differential Privacy retweeted
PETS@PET_Symposium11 Jul 2023
The 2023 Caspar Bowden PET Award winner is: "The 2020 Census Disclosure Avoidance System TopDown Algorithm" #pets23
1
10
2
34
12,306
Call for Papers - TPDP 2023 - Submission deadline July 7 differentialprivacy.org/tpdp…

Call for Papers - TPDP 2023 - Submission deadline July 7

The 9th Workshop on the Theory and Practice of Differential Privacy (TPDP 2023) will take place in Boston September 27-28, 2023. This is the first year the workshop is a standalone event. However,...

differentialprivacy.org
7
2
15
7,010
Open problem - Better privacy guarantees for larger groups differentialprivacy.org/open…

Open problem - Better privacy guarantees for larger groups

Consider a simple query counting the number of people in various mutually exclusive groups. In the differential privacy literature, it is typical to assume that each of these groups should be subject...

differentialprivacy.org
3
1
18
8,477
Composition Basics ift.tt/3L4oPfb

Composition Basics

Our data is subject to many different uses. Many entities will have access to our data and those entities will perform many different analyses that involve our data. The greatest risk to privacy is...

differentialprivacy.org
5
24
Load more