Data Provenance via Differential Auditing
This addresses data provenance auditing for machine learning practitioners, offering a more practical solution by removing the need for label information and training protocol knowledge, though it is incremental as it builds on existing auditing techniques.
The paper tackles the problem of auditing whether specific data was used to train a machine learning model, introducing DPDA, a framework that uses statistically significant differentials to distinguish training data without requiring shadow models or labeled outputs, and reports evaluations on real-world datasets showing its effectiveness.
Auditing Data Provenance (ADP), i.e., auditing if a certain piece of data has been used to train a machine learning model, is an important problem in data provenance. The feasibility of the task has been demonstrated by existing auditing techniques, e.g., shadow auditing methods, under certain conditions such as the availability of label information and the knowledge of training protocols for the target model. Unfortunately, both of these conditions are often unavailable in real applications. In this paper, we introduce Data Provenance via Differential Auditing (DPDA), a practical framework for auditing data provenance with a different approach based on statistically significant differentials, i.e., after carefully designed transformation, perturbed input data from the target model's training set would result in much more drastic changes in the output than those from the model's non-training set. This framework allows auditors to distinguish training data from non-training ones without the need of training any shadow models with the help of labeled output data. Furthermore, we propose two effective auditing function implementations, an additive one and a multiplicative one. We report evaluations on real-world data sets demonstrating the effectiveness of our proposed auditing technique.