{
  "id": "2206.08255",
  "version": "v1",
  "published": "2022-06-16T15:50:41.000Z",
  "updated": "2022-06-16T15:50:41.000Z",
  "title": "Gradient-Based Adversarial and Out-of-Distribution Detection",
  "authors": [
    "Jinsol Lee",
    "Mohit Prabhushankar",
    "Ghassan AlRegib"
  ],
  "comment": "International Conference on Machine Learning (ICML) Workshop on New Frontiers in Adversarial Machine Learning, July 2022",
  "categories": [
    "cs.LG",
    "cs.CV"
  ],
  "abstract": "We propose to utilize gradients for detecting adversarial and out-of-distribution samples. We introduce confounding labels -- labels that differ from normal labels seen during training -- in gradient generation to probe the effective expressivity of neural networks. Gradients depict the amount of change required for a model to properly represent given inputs, providing insight into the representational power of the model established by network architectural properties as well as training data. By introducing a label of different design, we remove the dependency on ground truth labels for gradient generation during inference. We show that our gradient-based approach allows for capturing the anomaly in inputs based on the effective expressivity of the models with no hyperparameter tuning or additional processing, and outperforms state-of-the-art methods for adversarial and out-of-distribution detection.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2022-06-16T15:50:41.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "out-of-distribution detection",
      "gradient-based adversarial",
      "gradient generation",
      "ground truth labels",
      "outperforms state-of-the-art methods"
    ],
    "tags": [
      "conference paper"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}