{
  "id": "1808.00087",
  "version": "v1",
  "published": "2018-07-31T22:13:39.000Z",
  "updated": "2018-07-31T22:13:39.000Z",
  "title": "Subsampled Rényi Differential Privacy and Analytical Moments Accountant",
  "authors": [
    "Yu-Xiang Wang",
    "Borja Balle",
    "Shiva Kasiviswanathan"
  ],
  "categories": [
    "cs.LG",
    "cs.CR",
    "stat.ML"
  ],
  "abstract": "We study the problem of subsampling in differential privacy (DP), a question that is the centerpiece behind many successful differentially private machine learning algorithms. Specifically, we provide a tight upper bound on the R\\'enyi Differential Privacy (RDP) (Mironov, 2017) parameters for algorithms that: (1) subsample the dataset, and then (2) apply a randomized mechanism M to the subsample, in terms of the RDP parameters of M and the subsampling probability parameter. This result generalizes the classic subsampling-based \"privacy amplification\" property of $(\\epsilon,\\delta)$-differential privacy that applies to only one fixed pair of $(\\epsilon,\\delta)$, to a stronger version that exploits properties of each specific randomized algorithm and satisfies an entire family of $(\\epsilon(\\delta),\\delta)$-differential privacy for all $\\delta\\in [0,1]$. Our experiments confirm the advantage of using our techniques over keeping track of $(\\epsilon,\\delta)$ directly, especially in the setting where we need to compose many rounds of data access.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2018-07-31T22:13:39.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "subsampled rényi differential privacy",
      "analytical moments accountant",
      "private machine learning algorithms",
      "differentially private machine learning"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}