Widespread Underestimation of Sensitivity in Differentially Private Libraries and How to Fix It
July 21, 2022 Β· Declared Dead Β· π Conference on Computer and Communications Security
"No code URL or promise found in abstract"
Evidence collected by the PWNC Scanner
Authors
SΓlvia Casacuberta, Michael Shoemate, Salil Vadhan, Connor Wagaman
arXiv ID
2207.10635
Category
cs.CR: Cryptography & Security
Citations
32
Venue
Conference on Computer and Communications Security
Last Checked
3 months ago
Abstract
We identify a new class of vulnerabilities in implementations of differential privacy. Specifically, they arise when computing basic statistics such as sums, thanks to discrepancies between the implemented arithmetic using finite data types (namely, ints or floats) and idealized arithmetic over the reals or integers. These discrepancies cause the sensitivity of the implemented statistics (i.e., how much one individual's data can affect the result) to be much larger than the sensitivity we expect. Consequently, essentially all differential privacy libraries fail to introduce enough noise to meet the requirements of differential privacy, and we show that this may be exploited in realistic attacks that can extract individual-level information from private query systems. In addition to presenting these vulnerabilities, we also provide a number of solutions, which modify or constrain the way in which the sum is implemented in order to recover the idealized or near-idealized bounds on sensitivity.
Community Contributions
Found the code? Know the venue? Think something is wrong? Let us know!
π Similar Papers
In the same crypt β Cryptography & Security
R.I.P.
π»
Ghosted
R.I.P.
π»
Ghosted
Membership Inference Attacks against Machine Learning Models
R.I.P.
π»
Ghosted
The Limitations of Deep Learning in Adversarial Settings
R.I.P.
π»
Ghosted
Practical Black-Box Attacks against Machine Learning
R.I.P.
π»
Ghosted
Distillation as a Defense to Adversarial Perturbations against Deep Neural Networks
R.I.P.
π»
Ghosted
Extracting Training Data from Large Language Models
Died the same way β π» Ghosted
R.I.P.
π»
Ghosted
Language Models are Few-Shot Learners
R.I.P.
π»
Ghosted
PyTorch: An Imperative Style, High-Performance Deep Learning Library
R.I.P.
π»
Ghosted
XGBoost: A Scalable Tree Boosting System
R.I.P.
π»
Ghosted