PROCHLO: Strong Privacy for Analytics in the Crowd

By Ú. Erlingsson, Petros Maniatis, Ilya Mironov, David Lie, Mitch Rudominer, Julien Tinnés. 2017

How researchers see this paper

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

In Privacy & Security

Citations in the corpus

Privacy Amplification via Random Check-Ins (2020, methodology): “For this reason, DP has been adopted by companies including Google [20, 10, 18], Apple [2], Microsoft [13], and LinkedIn [31], as well as the US Census Bureau [26].”
Auditing Differentially Private Machine Learning: How Private is Private SGD? (2020, background): “Diﬀerential privacy [DMNS06] has become the de facto standard for guaranteeing privacy in machine learning and statistical analysis, and is now being deployed by many organizations including Apple [TVV + 17], Google [EPK14, BEM + 17, PSM + 18], and the US Census Bureau [HMA + 17].”
Pure Differentially Private Summation from Anonymous Messages (2020, methodology): “While an analogous setup was ﬁrst introduced in crytpography by Ishai et al. in their work on cryptography from anonymity [IKOS06], the shuﬄed model was ﬁrst proposed for privacy-preserving computations by Bittau et al. [BEM + 17] in their Encode-Shuﬄe-Analyze architecture.”
Improved Summation from Shuffling (2019, background): “The work of Ishai et al. is of signiﬁcant relevance in the context of the recently proposed shuﬄe model of Diﬀerential Privacy (DP) [2] (see also [4, 3]).”
Scalable and Differentially Private Distributed Aggregation in the Shuffled Model (2019, methodology): “In contrast, protocols based on shufﬂing work with much weaker assumptions on the server [5, 7].”
Amplification by Shuffling: From Local to Central Differential Privacy via Anonymity (2018, background): “Our direct motivation is the Encode, Shufﬂe, Analyze (ESA) architecture and P ROCHLO implementation of Bittau et al. [BEM + 17], which relies on an explicit intermediary that processes LDP reports from users to ensure their anonymity.”
Differentially-Private "Draw and Discard" Machine Learning (2018, background): “…private applications and the very recent work of [19]; for the second – from the similarities between DDML and the shufﬂing strategy of [6] and privacy ampliﬁcation by sampling exploited by [32], [47], [1]; for the third – from recent work in differentially private machine learning…”
Encode, Shuffle, Analyze Privacy Revisited: Formalizations and Empirical Evaluation (2020, background): “In addition to the above, as in the ESA [2] architecture, an intermediary termed the shufﬂer can be used to ensure the anonymity of reports without having visibility into report contents (thanks to cryptography).”
SGX-LKL: Securing the Host OS Interface for Trusted Execution (2019, background): “Bittau et al. [13] introduce an oblivious architecture for monitoring client software behaviour while protecting user privacy.”
Pan-Private Uniformity Testing (2019, background): “Rigorous privacy guarantees have driven increasing adoption of differential privacy by industry [8, 10, 21, 29], government [1]), and academic researchers [37, 34].”
Gradient-Leaks: Understanding and Controlling Deanonymization in Federated Learning (2018, background): “Such a deanonymization objective undermines sanitization mechanisms which de-identify model updates, such as decoupling the update from user identity [15], stripping away identiﬁable metadata [35,83], and blind-shufﬂing mechanisms [14].”
Practical and Robust Privacy Amplification with Multi-Party Differential Privacy (2019, methodology): “The shufﬂing idea was originally proposed in Prochlo [12], where a shufﬂer is inserted between the users and the server to break the linkage between the report and the user identi-ﬁcation.”
Differentially Private Set Union (2020, background): “Diﬀerential privacy is deployed in many industries, including Apple Apple [2017], Google Erlingsson et al. [2014], Bittau et al. [2017], Microsoft Ding et al. [2017], Mozilla Avent et al. [2017], and the US Census Bureau Abowd [2016], Kuo et al. [2018].”
Locally Differentially Private Frequent Itemset Mining (2018, methodology): “For example, researchers from Google developed RAPPOR [18], [20] and Prochlo [8], which are included as part of Chrome.”
CoinPress: Practical Private Mean and Covariance Estimation (2020, unknown): “A large body of work now shows that, in principle, nearly every statistical task can be solved privately, and differential privacy is now being deployed by Apple [Dif17], Google [EPK14, BEM + 17], Microsoft [DKY17], and the US Census Bureau [DLS + 17].”
Privacy and Utility Preserving Sensor-Data Transformations (2019, background): “…mechanisms [9] provide plausible deniability guarantees when estimating, for example, the mean or frequency of a common variable among users [10], with multi-dimensional data released sequentially a more practical privacy model is inferential privacy [11, 12, 13] that measures the diﬀerence…”
Locally Private k-Means in One Round (2021, methodology): “To demonstrate this, we give an algorithm in the shuffle model (Bittau et al., 2017; Erlingsson et al., 2019; Cheu et al., 2019)2 with a similar approximation ratio but with a smaller additive error of Õ(kO↵(1) · p d/✏).”
An Algorithmic Framework For Differentially Private Data Analysis on Trusted Processors (2018, methodology): “As we mentioned earlier, off-the-shelf DP-algorithms may not be suitable for TEEs for two main reasons.”
Local Differential Privacy for Evolving Data (2018, methodology): “Our method improves over the naïve solution of simply recomputing the statistic every epoch – which would lead to either privacy parameter or error that scales linearly with the (1)Although the value of that Apple guarantees over the course of say, a week, is not meaningful on its own, Apple does ta”