• DocumentCode
    2857088
  • Title

    Two-source extractors for leaky sources

  • Author

    Yu Yu ; Xiangxue Li ; Haifeng Qian

  • Author_Institution
    Inst. for Interdiscipl. Inf. Sci., Tsinghua Univ., Beijing, China
  • fYear
    2012
  • fDate
    3-7 Sept. 2012
  • Firstpage
    452
  • Lastpage
    456
  • Abstract
    A (worst-case) 2-source extractor is a deterministic algorithm that transforms pairwise independent weak random sources into almost uniform random strings. Despite non-constructive proofs that such objects exist with almost optimal parameters, it has been a longstanding open problem to construct `explicit´ (aka efficient) functions for sources of `small´ constant entropy rate. In particular, best known constructions either require entropy rate of at least 0.4999 (due to Bourgain), or one source must remain with constant entropy rate above half (due to Raz). Motivated by cryptographic applications, we observe that if one source is a leaky source (or it contains a few deterministically extractable entropy), then we will be able to efficiently extract almost all entropy from both sources with nearly optimal entropy loss. Further, our extractor (for leaky sources) does not suffer from the half entropy rate barrier, and it works for all linear (and even sub-linear) entropy sources. The extractor is constructed using the technique of alternating extraction by Dziembowski and Pietrzak (FOCS 2007). Finally, we show that the extractor is almost a worse-case extractor (for the same parameters) in the sense that it only fails for a negligible fraction of sources.
  • Keywords
    cryptography; constant entropy rate; cryptographic applications; deterministic algorithm; leaky sources; optimal parameters; two source extractors; uniform random strings; Computer science; Conferences; Cryptography; Data mining; Entropy; Helium; Information theory;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Information Theory Workshop (ITW), 2012 IEEE
  • Conference_Location
    Lausanne
  • Print_ISBN
    978-1-4673-0224-1
  • Electronic_ISBN
    978-1-4673-0222-7
  • Type

    conf

  • DOI
    10.1109/ITW.2012.6404713
  • Filename
    6404713