On significance of the least significant bits for differential privacy

By Ilya Mironov. 2012

248 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Information Systems, 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 Artificial Intelligence

In Information Systems

In Privacy & Security

Citations in the corpus

Advances and Open Problems in Federated Learning (2019, neutral): “, [296, 192] for differential privacy).”
Honeycrisp: large-scale differentially private aggregation without a trusted core (2019, background): “Thus, the noise generation must be carefully implemented (for instance, with a snapping mechanism as described in [58]) to ensure differential privacy and to prevent such attacks.”
Privacy Loss in Apple's Implementation of Differential Privacy on MacOS 10.12 (2017, background): “It has been observed that properly implementing algorithms claiming to preserve differential privacy is non-trivial in practice [4, 5, 13].”
RAPPOR: Randomized Aggregatable Privacy-Preserving Ordinal Response (2014, background): “Assuming absolute compliance with the randomization protocol (an assumption that may not hold for human subjects, and can even be nontrivial for algorithmic implementations [23]), it is easy to see that in a case where both “Yes” and “No” answers can be denied (flipping two fair coins), the true num”
Signal Processing and Machine Learning with Differential Privacy (2013, background): “In particu lar, much of the theory has been developed for data taking discrete values, and there are many challenges raised by con tinuous data, ranging from the implementation of differen tially private algorithms [20] to theoretical foundations [21].”
The Algorithmic Foundations of Differential Privacy (2014, background): “The vulnerability of differential privacy to inappropriate implementations of floating point numbers was observed by Mironov, who proposed a mitigation [63].”
The optimal mechanism in differential privacy (2012, unknown): “, [10], [11], [12], [13], [14], [5], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46].”
An Algorithmic Framework For Differentially Private Data Analysis on Trusted Processors (2018, background): “Memory side-channels should not be confused with vulnerabilities introduced by ﬂoating-point implementations [33].”
Achieving Differential Privacy in Secure Multiparty Data Aggregation Protocols on Star Networks (2017, background): “In fact, textbook ﬂoating point implementations are vulnerable to attacks as demonstrated by Mironov [17] while integer or ﬁxed-point representations are not vulnerable to such attacks.”
Differentially Private SQL with Bounded User Contribution (2019, background): “Some of these issues have previously been explored in the literature, notably regarding the consequences of using a floating-point representation [35] with Laplace noise.”
Differentially private data aggregation with optimal utility (2014, methodology): “This mismatch has been shown to give rise to several attacks, as shown by Mironov [45] and Gazeau et al.”
EKTELO: A Framework for Defining Differentially-Private Computations (2018, methodology): “Initialize ✏ktelo 2: D Where(D, sex == ‘M’ AND age 2 [30, 39]) .”
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.”
Efficient Differentially Private F0 Linear Sketching (2020, background): “Discretizing a real-valued private mechanism is non-trivial: Without sufficient care, one might lose privacy due to rounding in an implementation, as argued by Mironov [36].”