Optimal private halfspace counting via discrepancy

Shanmugavelayutham Muthukrishnan, Aleksandar Nikolov

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Abstract

    A range counting problem is specified by a set P of size |P| = n of points in ℝ d, an integer weight x p associated to each point p ε P, and a range space R ⊆ 2 P. Given a query range R ε R, the output is R(x) = Σ pεR x p. The average squared error of an algorithm A is 1/|R|Σ RεR((A(R, x) - R(x))) 2. Range counting for different range spaces is a central problem in Computational Geometry. We study (ε, δ)-differentially private algorithms for range counting. Our main results are for the range space given by hyperplanes, that is, the halfspace counting problem. We present an (ε, δ)-differentially private algorithm for halfspace counting in d dimensions which is O(n 1-1/d) approximate for average squared error. This contrasts with the Ω(n) lower bound established by the classical result of Dinur and Nissim on approximation for arbitrary subset counting queries. We also show a matching lower bound of Ω(n 1-1/d) approximation for any (ε, δ)-differentially private algorithm for halfspace counting. Both bounds are obtained using discrepancy theory. For the lower bound, we use a modified discrepancy measure and bound approximation of (ε, δ)-differentially private algorithms for range counting queries in terms of this discrepancy. We also relate the modified discrepancy measure to classical combinatorial discrepancy, which allows us to exploit known discrepancy lower bounds. This approach also yields a lower bound of Ω((log n) d-1) for (ε, δ)-differentially private orthogonal range counting in d dimensions, the first known superconstant lower bound for this problem. For the upper bound, we use an approach inspired by partial coloring methods for proving discrepancy upper bounds, and obtain (ε, δ)-differentially private algorithms for range counting with polynomially bounded shatter function range spaces.

    Original languageEnglish (US)
    Title of host publicationSTOC '12 - Proceedings of the 2012 ACM Symposium on Theory of Computing
    Pages1285-1292
    Number of pages8
    DOIs
    StatePublished - Jun 26 2012
    Event44th Annual ACM Symposium on Theory of Computing, STOC '12 - New York, NY, United States
    Duration: May 19 2012May 22 2012

    Publication series

    NameProceedings of the Annual ACM Symposium on Theory of Computing
    ISSN (Print)0737-8017

    Other

    Other44th Annual ACM Symposium on Theory of Computing, STOC '12
    CountryUnited States
    CityNew York, NY
    Period5/19/125/22/12

    Fingerprint

    Computational geometry
    Coloring

    Keywords

    • combinatorial discrepancy
    • differential privacy
    • range counting

    ASJC Scopus subject areas

    • Software

    Cite this

    Muthukrishnan, S., & Nikolov, A. (2012). Optimal private halfspace counting via discrepancy. In STOC '12 - Proceedings of the 2012 ACM Symposium on Theory of Computing (pp. 1285-1292). (Proceedings of the Annual ACM Symposium on Theory of Computing). https://doi.org/10.1145/2213977.2214090

    Optimal private halfspace counting via discrepancy. / Muthukrishnan, Shanmugavelayutham; Nikolov, Aleksandar.

    STOC '12 - Proceedings of the 2012 ACM Symposium on Theory of Computing. 2012. p. 1285-1292 (Proceedings of the Annual ACM Symposium on Theory of Computing).

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Muthukrishnan, S & Nikolov, A 2012, Optimal private halfspace counting via discrepancy. in STOC '12 - Proceedings of the 2012 ACM Symposium on Theory of Computing. Proceedings of the Annual ACM Symposium on Theory of Computing, pp. 1285-1292, 44th Annual ACM Symposium on Theory of Computing, STOC '12, New York, NY, United States, 5/19/12. https://doi.org/10.1145/2213977.2214090
    Muthukrishnan S, Nikolov A. Optimal private halfspace counting via discrepancy. In STOC '12 - Proceedings of the 2012 ACM Symposium on Theory of Computing. 2012. p. 1285-1292. (Proceedings of the Annual ACM Symposium on Theory of Computing). https://doi.org/10.1145/2213977.2214090
    Muthukrishnan, Shanmugavelayutham ; Nikolov, Aleksandar. / Optimal private halfspace counting via discrepancy. STOC '12 - Proceedings of the 2012 ACM Symposium on Theory of Computing. 2012. pp. 1285-1292 (Proceedings of the Annual ACM Symposium on Theory of Computing).
    @inproceedings{45a9a4852bb0400d935c1aa6b4e15970,
    title = "Optimal private halfspace counting via discrepancy",
    abstract = "A range counting problem is specified by a set P of size |P| = n of points in ℝ d, an integer weight x p associated to each point p ε P, and a range space R ⊆ 2 P. Given a query range R ε R, the output is R(x) = Σ pεR x p. The average squared error of an algorithm A is 1/|R|Σ RεR((A(R, x) - R(x))) 2. Range counting for different range spaces is a central problem in Computational Geometry. We study (ε, δ)-differentially private algorithms for range counting. Our main results are for the range space given by hyperplanes, that is, the halfspace counting problem. We present an (ε, δ)-differentially private algorithm for halfspace counting in d dimensions which is O(n 1-1/d) approximate for average squared error. This contrasts with the Ω(n) lower bound established by the classical result of Dinur and Nissim on approximation for arbitrary subset counting queries. We also show a matching lower bound of Ω(n 1-1/d) approximation for any (ε, δ)-differentially private algorithm for halfspace counting. Both bounds are obtained using discrepancy theory. For the lower bound, we use a modified discrepancy measure and bound approximation of (ε, δ)-differentially private algorithms for range counting queries in terms of this discrepancy. We also relate the modified discrepancy measure to classical combinatorial discrepancy, which allows us to exploit known discrepancy lower bounds. This approach also yields a lower bound of Ω((log n) d-1) for (ε, δ)-differentially private orthogonal range counting in d dimensions, the first known superconstant lower bound for this problem. For the upper bound, we use an approach inspired by partial coloring methods for proving discrepancy upper bounds, and obtain (ε, δ)-differentially private algorithms for range counting with polynomially bounded shatter function range spaces.",
    keywords = "combinatorial discrepancy, differential privacy, range counting",
    author = "Shanmugavelayutham Muthukrishnan and Aleksandar Nikolov",
    year = "2012",
    month = "6",
    day = "26",
    doi = "10.1145/2213977.2214090",
    language = "English (US)",
    isbn = "9781450312455",
    series = "Proceedings of the Annual ACM Symposium on Theory of Computing",
    pages = "1285--1292",
    booktitle = "STOC '12 - Proceedings of the 2012 ACM Symposium on Theory of Computing",

    }

    TY - GEN

    T1 - Optimal private halfspace counting via discrepancy

    AU - Muthukrishnan, Shanmugavelayutham

    AU - Nikolov, Aleksandar

    PY - 2012/6/26

    Y1 - 2012/6/26

    N2 - A range counting problem is specified by a set P of size |P| = n of points in ℝ d, an integer weight x p associated to each point p ε P, and a range space R ⊆ 2 P. Given a query range R ε R, the output is R(x) = Σ pεR x p. The average squared error of an algorithm A is 1/|R|Σ RεR((A(R, x) - R(x))) 2. Range counting for different range spaces is a central problem in Computational Geometry. We study (ε, δ)-differentially private algorithms for range counting. Our main results are for the range space given by hyperplanes, that is, the halfspace counting problem. We present an (ε, δ)-differentially private algorithm for halfspace counting in d dimensions which is O(n 1-1/d) approximate for average squared error. This contrasts with the Ω(n) lower bound established by the classical result of Dinur and Nissim on approximation for arbitrary subset counting queries. We also show a matching lower bound of Ω(n 1-1/d) approximation for any (ε, δ)-differentially private algorithm for halfspace counting. Both bounds are obtained using discrepancy theory. For the lower bound, we use a modified discrepancy measure and bound approximation of (ε, δ)-differentially private algorithms for range counting queries in terms of this discrepancy. We also relate the modified discrepancy measure to classical combinatorial discrepancy, which allows us to exploit known discrepancy lower bounds. This approach also yields a lower bound of Ω((log n) d-1) for (ε, δ)-differentially private orthogonal range counting in d dimensions, the first known superconstant lower bound for this problem. For the upper bound, we use an approach inspired by partial coloring methods for proving discrepancy upper bounds, and obtain (ε, δ)-differentially private algorithms for range counting with polynomially bounded shatter function range spaces.

    AB - A range counting problem is specified by a set P of size |P| = n of points in ℝ d, an integer weight x p associated to each point p ε P, and a range space R ⊆ 2 P. Given a query range R ε R, the output is R(x) = Σ pεR x p. The average squared error of an algorithm A is 1/|R|Σ RεR((A(R, x) - R(x))) 2. Range counting for different range spaces is a central problem in Computational Geometry. We study (ε, δ)-differentially private algorithms for range counting. Our main results are for the range space given by hyperplanes, that is, the halfspace counting problem. We present an (ε, δ)-differentially private algorithm for halfspace counting in d dimensions which is O(n 1-1/d) approximate for average squared error. This contrasts with the Ω(n) lower bound established by the classical result of Dinur and Nissim on approximation for arbitrary subset counting queries. We also show a matching lower bound of Ω(n 1-1/d) approximation for any (ε, δ)-differentially private algorithm for halfspace counting. Both bounds are obtained using discrepancy theory. For the lower bound, we use a modified discrepancy measure and bound approximation of (ε, δ)-differentially private algorithms for range counting queries in terms of this discrepancy. We also relate the modified discrepancy measure to classical combinatorial discrepancy, which allows us to exploit known discrepancy lower bounds. This approach also yields a lower bound of Ω((log n) d-1) for (ε, δ)-differentially private orthogonal range counting in d dimensions, the first known superconstant lower bound for this problem. For the upper bound, we use an approach inspired by partial coloring methods for proving discrepancy upper bounds, and obtain (ε, δ)-differentially private algorithms for range counting with polynomially bounded shatter function range spaces.

    KW - combinatorial discrepancy

    KW - differential privacy

    KW - range counting

    UR - http://www.scopus.com/inward/record.url?scp=84862589671&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84862589671&partnerID=8YFLogxK

    U2 - 10.1145/2213977.2214090

    DO - 10.1145/2213977.2214090

    M3 - Conference contribution

    AN - SCOPUS:84862589671

    SN - 9781450312455

    T3 - Proceedings of the Annual ACM Symposium on Theory of Computing

    SP - 1285

    EP - 1292

    BT - STOC '12 - Proceedings of the 2012 ACM Symposium on Theory of Computing

    ER -