Differential Privacy Under Fire

By Andreas Haeberlen, Benjamin C. Pierce, Arjun Narayan. 2011

213 citations recorded by the source corpus.

Subfields: Safety, Robustness & Interpretability, Privacy & Security

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Safety, Robustness & Interpretability

In Privacy & Security

Citations in the corpus

Advances and Open Problems in Federated Learning (2019, neutral): “, [296, 192] for differential privacy).”
Deep Models Under the GAN: Information Leakage from Collaborative Deep Learning (2017, background): “Covert channels, however, can be used to defeat DP-protected databases as shown in [36].”
Honeycrisp: large-scale differentially private aggregation without a trusted core (2019, methodology): “We can enable this by writing the queries in a language such as Fuzz [39], which comes with a static analysis that bounds the sensitivity.”
Towards Statistical Queries over Distributed Private User Data (2012, background): “In spite of the fact that an increasing number of queries can lead to increased privacy loss, we find differential privacy to be an attractive model both because it does provide measurable privacy, and because there is substantial ongoing effort in developing its uses [32, 38, 39, 40, 47] and unders”
Demystifying Membership Inference Attacks in Machine Learning as a Service (2019, unknown): “Machine learning-as-a-service APIs may additionally be hardened through limiting queries and/or allocating a differential privacy budget to query responses [43].”
DJoin: Differentially Private Join Queries over Distributed Databases (2012, unknown): “Differentially private query processors: PINQ [23], Airavat [32], and Fuzz [16] are query processors that support differential privacy, but they assume a centralized setting in which a single entity has access to the entire data.”
GUPT: privacy preserving data analysis made easy (2012, unknown): “[10] identified three possible side-channel attacks against differentially private systems.”
Differential Privacy (2015, background): “The implementation of PINQ has a number of side-channels as pointed out in [12]; these and other implementation weaknesses ([19]) are certainly present in our implementation and it has not been our goal to focus on those.”
Satisfiability modulo counting: a new approach for analyzing privacy properties (2014, background): “However, implementing programs that satisfy these deﬁnitions is notoriously difﬁcult, as information leaks may arise from unexpected sources [26, 36]: exceptional control ﬂow, the granularity of ﬂoating-point numbers, and even the decision not to publish a result can give an adversary enough…”
DStress: Efficient Differentially Private Computations on Distributed Data (2017, methodology): “, using linear type systems [31, 38] or a system like CertiPriv [4].”
On significance of the least significant bits for differential privacy (2012, background): “The distribution of ˜ f ϵ on two inputs where f ( D ) = 0 and f ( D ′ ) = 1 is plotted in Figure 1.”
A differential privacy protecting K-means clustering algorithm based on contour coefficients (2018, methodology): “D and D’ are two neighboring subsets between which the difference is no more than one record. ε is a specified constant and is called privacy protection budget[13].”
Practical Differential Privacy for SQL Queries Using Elastic Sensitivity (2017, background): “Fuzz [20,21] enforces differential privacy for functional programs, but does not support one-to-many or many-to-many joins.”
Differential Privacy by Typing in Security Protocols (2013, background): “[35] and propose alternative notions of privacy for queries on statistical databases, such as relaxations of differential privacy [36], noiseless definitions [37], and zero-knowledge based definitions for social networks [38].”
Edge Computing Security: State of the Art and Challenges (2019, unknown): “[93].”
Distributed Differential Privacy and Applications (2015, methodology): “VerDP’s query language is based on the Fuzz language for differentially private data analysis (Reed and Pierce, 2010; Haeberlen et al., 2011; Gaboardi et al., 2013) because: 1) it is sufficiently expressive to implement a number of practical differentially-private queries, and 2) it uses static anal”