{
  "id": "1911.03030",
  "version": "v1",
  "published": "2019-11-08T03:57:41.000Z",
  "updated": "2019-11-08T03:57:41.000Z",
  "title": "Certified Data Removal from Machine Learning Models",
  "authors": [
    "Chuan Guo",
    "Tom Goldstein",
    "Awni Hannun",
    "Laurens van der Maaten"
  ],
  "comment": "Submitted to AISTATS 2020",
  "categories": [
    "cs.LG",
    "stat.ML"
  ],
  "abstract": "Good data stewardship requires removal of data at the request of the data's owner. This raises the question if and how a trained machine-learning model, which implicitly stores information about its training data, should be affected by such a removal request. Is it possible to \"remove\" data from a machine-learning model? We study this problem by defining certified removal: a very strong theoretical guarantee that a model from which data is removed cannot be distinguished from a model that never observed the data to begin with. We develop a certified-removal mechanism for linear classifiers and empirically study learning settings in which this mechanism is practical.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2019-11-08T03:57:41.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "machine learning models",
      "certified data removal",
      "machine-learning model",
      "datas owner",
      "implicitly stores information"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}