Data Privacy: From Foundations to Applications (Simons Institute workshop, 2019)
2019-03-04..2019-03-08
Berkeley, United States
who it's for
privacy researchers, differential privacy engineers, government statistical agency staff, data stewards at public institutions, applied cryptography engineers, data scientists working with sensitive data, legal and compliance officers, privacy tool developers
topics
differential privacy mechanisms and deployment; TopDown census microdata pipeline; smooth sensitivity for k-nearest neighbor; differentially private synthetic data; per-user privacy budget accounting; relational type verification for DP; automated DP verification with SMT; bayesian optimization of privacy-utility tradeoffs; formalizing singling out and GDPR concepts; PSI: DP releases for research data; secure multi-party computation and Conclave; maximum observed sensitivity for regression
| title | author | length |
|---|---|---|
| Stepping-up: The Census Bureau Sets an Example of How to Be a Good Data Steward in the 21st Century | John Abowd | 01h 00m |
Publicly disclosed experiments reconstruct all 308,745,538 2010 person records from five variables and match block/sex/age/race/ethnicity exactly for 46% (71%… | ||
| Consistency with External Knowledge: The TopDown Algorithm | Dan Kifer | 45m |
TopDown builds a differentially private microdata pipeline for the 2020 census that publishes hierarchical histograms over 12 billion cells (about 300M people… | ||
| Smooth Sensitivity for k-Nearest Neighbor | Christopher Clifton | 48m |
Proposes using smooth sensitivity for k‑nearest‑neighbor imputation to avoid the astronomic global sensitivity that simple donor reuse causes, derives an upper… | ||
| Hybrid Legal-Technical Concepts of Privacy | Alexandra Wood | 44m |
Argues that common legal concepts like HIPAA safe harbor fail empirically, builds a practical catalog of controls and a tiered access architecture that uses… | ||
| Quantifying Privacy in Law | Felix Wu | 45m |
Law treats privacy more as enforceable violations than a single formal value and is willing to quantify some things (the HIPAA safe harbor removes 18… | ||
| Discovery, Replication and Reuse of Sensitive Scientific Data with PSI | James Honaker | 48m |
A deployable system called PSI sits on Dataverse and generates differentially private summary releases for sensitive repositories, with a budgeter GUI that… | ||
| Modular Approach to Cloud Security | Mayank Varia | 44m |
Argues that security must be proven compositionally and shows a working stack: formal universally composable security proofs for kernels, hardware, OpenStack… | ||
| Learning the Privacy-Utility Trade-off with Bayesian Optimization | Borja Balle | 46m |
Implementation uses multi-objective Bayesian optimization with a Gaussian process surrogate and a hypervolume probability of improvement acquisition to map… | ||
| Formal Modelling of Information-flow Control Systems: Some Lessons Learned | Dave Sands | 48m |
Proposes replacing a global privacy budget with per-user privacy budgets and implements a dynamic enforcement system that tracks provenance-tracking for every… | ||
| Proving Differential Privacy via Relational Types | Danfeng Zhang | 45m |
A relational dependent type system (implemented as LightDP and extended as ShadowDP) converts probabilistic programs into non‑deterministic targets that track… | ||
| Automated Verification meets Differential Privacy | Aws Albarghouthi | 45m |
Automatic verifier transforms sampling into assume statements, encodes failure probability as a program variable, and hands the result to off‑the‑shelf SMT/SAT… | ||
| A Practical Method to Reduce Privacy Loss when Disclosing Statistics Based on Small Samples | John Friedman | 45m |
Releases local regression predictions with added noise using a data-driven sensitivity cap called maximum observed sensitivity (MOS) (Chi scales as Chi/n)… | ||
| From Health Data to Location Traces: Addressing Spatiotemporal Correlations in Differential Privacy | Li Xiong | 46m |
Two projects: building differentially private clinical registries and protecting single-user location traces. For EHRs, tensor factorization runs in a… | ||
| Challenges in Scaling Privacy Enhancing Technologies to Campus Level IoT Testbed | Sharad Mehrotra | 45m |
Campus IoT testbed TIPPERS runs in 35 buildings and ingests roughly ~1M location events/day (≈44k/day per building), integrates Pegasus for differential… | ||
| Practical Experience with Making Synthetic Data Differentially Private | Gillian Raab | 45m |
Practical work tests three concrete ways to make synthetic tabular data differentially private: Laplace on full table, noisy marginals + IPF, and MWEM… | ||
| Predicate Singling Out: Towards Formalizing the GDPR Notion of Singling Out | Aloni Cohen | 45m |
Formalizes GDPR “singling out” as Predicate Singling Out (PSO): an adversary outputs a predicate whose weight (probability an individual satisfies it) is tiny… | ||
| What All We've Been Up To at Knexus Research | Christine Task | 39m |
Knexus Research presents three concrete efforts: a DARPA/Brandeis crisis detector that builds a social‑network generator and synthetic caller data, develops… | ||
| Accuracy Disparities and Social Choices in the Design of Privacy Mechanisms | Gerome Miklau | 44m |
Claims that common differential privacy pipelines produce uneven harms: applying Laplace mechanism to census-style counts yields unbiased equal-variance noise… | ||
| Differential Privacy from a Statistical Perspective–Obtaining Valid Inferences... | Jörg Drechsler | 47m |
Argues that releasing differentially private synthetic microdata via a full cross‑classification with independent two‑sided geometric mechanism noise yields… | ||
| pMSE Mechanism: Differentially Private Synthetic Data with Maximal Distributional Similarity | Joshua Snoke | 41m |
Introduces a mechanism that uses the exponential mechanism to sample model parameters and then generates differentially private synthetic datasets, optimizing… | ||
| Deploying DP: Experiences from the Front Lines and Open Challenges | Ashwin Machanavajjhala | 47m |
Worked with the US Census’s OnTheMap product, formalizes Title 13 requirements into a pufferfish framework that defines which relational secrets must be… | ||
| Statistical Disclosure Limitation: Where do we go from here? | Natalie Shlomo | 49m |
Statistical agencies are moving established table builders toward non-interactive table builders that fix the release space in advance and apply perturbation… | ||
| Matching Uses and Data Protections for Government Data | Micah Altman | 41m |
Presents an empirical catalog of how government data is disseminated (published estimates, interactive queries, dynamic maps, public use files, restricted… | ||
| Some Applications of Contextual Integrity | Helen Nissenbaum | 41m |
Contextual integrity defines privacy as the appropriate flow of information using five CI parameters (sender, recipient, subject, attribute, transmission… | ||
| DARPA and Privacy Research | Joshua Baron | 43m |
DARPA funds Brandeis prototypes and a new RACE effort that produce concrete systems: an open-source PALS Android OS enforces fine-grain app data access; Prio… | ||
| PANEL: Where Are Statistical Agencies Going? | Jörg Drechsler, Frauke Kreuter, John Abowd, Natalie Shlomo | 55m |
Panel reports that traditional suppression and record‑swapping are revealed as vulnerable to reconstruction, prompting adoption of formal methods: Census… | ||
| Making Privacy Technology Accessible: Benchmarks and Platforms | Michael Hay | 45m |
Empirical benchmark DPBench/DP Comp shows that contradictory paper results trace to signal strength (epsilon, record count) rather than universal algorithm… | ||
| Towards Privacy-Respectful Cyber-Physical Systems: Two Perspectives | Ruoxi Jia | 38m |
Describes two concrete approaches: a data-valuation pipeline that assigns payments using Shapley value for ML, with an exact k-NN scheme that computes values… | ||
| What We’re Doing in Australia – An API for Confidentialized Contingency Tables | Arthur Street | 29m |
API lets authorized users generate custom contingency tables (counts, plus means/sums where available) against ABS unit‑record data with PII removed, currently… | ||
| Relational Symbolic Execution for Differential Privacy | Gian Pietro Farina | 24m |
Proposes relational symbolic execution that runs paired symbolic traces and encodes Laplace sampling with an axiom that consumes an epsilon budget, turning… | ||
| Privately Learning High-Dimensional Distributions | Gautam Kamath | 36m |
Claims a private algorithm that learns d‑dimensional Gaussians in total variation with sample complexity near d^2/α^2 plus d^2/(αε) and an extra… | ||
| Open DP: A Proposal for an Open-Source Suite of Differential Privacy Tools | Salil Vadhan | 47m |
Proposal to build an open-source, community‑managed suite called OpenDP that supplies a library of vetted differentially private algorithms, APIs for both… | ||
| nothing matches | ||