On the Privacy Risks of Model Explanations
Open accessThe DOI matches its Crossref record; 28 of the 28 references it cites were found in a registry or an index.
Abstract
Privacy and transparency are two key foundations of trustworthy machine learning. Model explanations offer insights into a model's decisions on input data, whereas privacy is primarily concerned with protecting information about the training data. We analyze connections between model explanations and the leakage of sensitive information about the model's training set. We investigate the privacy risks of feature-based model explanations using membership inference attacks: quantifying how much model predictions plus their explanations leak information about the presence of a datapoint in the training set of a model. We extensively evaluate membership inference attacks based on feature-based model explanations, over a variety of datasets. We show that backpropagation-based explanations can leak a significant amount of information about individual training datapoints. This is because they reveal statistical information about the decision boundaries of the model about an input, which can reveal its membership. We also empirically investigate the trade-off between privacy and explanation quality, by studying the perturbation-based model explanations.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155
Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps
Simonyan, K., Vedaldi, A., & Zisserman, A. (2013) · arXiv · doi:10.48550/arxiv.1312.6034 · Cited by 5,370
Artificial intelligence in healthcare: past, present and future
Jiang, F., Jiang, Y., Zhi, H., Dong, Y., Li, H., Ma, S., Wang, Y., Dong, Q., Shen, H., & Wang, Y. (2017) · Stroke and Vascular Neurology · doi:10.1136/svn-2017-000101 · Cited by 4,837
Membership Inference Attacks Against Machine Learning Models
Shokri, R., Stronati, M., Song, C., & Shmatikov, V. (2017) · 2017 IEEE Symposium on Security and Privacy (SP) · doi:10.1109/sp.2017.41 · Cited by 4,754 · cited 5×
“sary leverage transparency reports in order to infer private information? We focus on inferring the presence of individual data points in training set of the model, using membership inference attacks [27] and reconstruction attacks. We analyze feature-based explanation algorithms, with the emphasis on gradient-based m…”
BackgroundMethodInfluentialSelf-citationOn Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation
Bach, S., Binder, A., Montavon, G., Klauschen, F., Müller, K.-R., & Samek, W. (2015) · PLOS ONE · doi:10.1371/journal.pone.0130140 · Cited by 4,524 · cited 1×
“We analyze both backpropagation-based model explanations, with an emphasis on gradient-based methods (Baehrens et al. 2009; Klauschen et al. 2015; Shrikumar, Greenside, and Kundaje 2017b; Sliwinski, Strobel, and Zick 2019; Sundararajan, Taly, and Yan 2017) and perturbation-based methods (Ribeiro, Singh, and Guestrin 2…”
MethodStriving for Simplicity: The All Convolutional Net
Springenberg, J. T., Dosovitskiy, A., Brox, T., & Riedmiller, M. (2014) · arXiv · doi:10.48550/arxiv.1412.6806 · Cited by 3,447
Understanding Machine Learning: From Theory to Algorithms
Shalev-Shwartz, S., & Ben-David, S. (2014) · doi:10.1017/cbo9781107298019 · Cited by 3,115
Learning Important Features Through Propagating Activation Differences
Shrikumar, A., Greenside, P., & Kundaje, A. (2017) · arXiv · doi:10.48550/arxiv.1704.02685 · Cited by 2,761
Anchors: High-Precision Model-Agnostic Explanations
Ribeiro, M. T., Singh, S., & Guestrin, C. (2018) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v32i1.11491 · Cited by 2,162 · cited 1×
“that the model is a neural network [6, 42, 53], or that we know the source code [14, 37]).”
BackgroundEuropean Union Regulations on Algorithmic Decision Making and a “Right to Explanation”
Goodman, B., & Flaxman, S. (2017) · AI Magazine · doi:10.1609/aimag.v38i3.2741 · Cited by 2,131 · cited 2×
“[19] B Goodman and S R Flaxman.”
BackgroundUnderstanding Black-box Predictions via Influence Functions
Koh, P. W., & Liang, P. (2017) · arXiv · doi:10.48550/arxiv.1703.04730 · Cited by 1,514 · cited 1×
“We use the modifed version described in (Koh and Liang 2017) where each patient has 127 features which are demographic (e.”
MethodSmoothGrad: removing noise by adding noise
Smilkov, D., Thorat, N., Kim, B., Viégas, F., & Wattenberg, M. (2017) · arXiv · doi:10.48550/arxiv.1706.03825 · Cited by 1,467 · cited 8×
“(4)We test SmoothGrad [42], as well as LIME [32]; however the latter takes relatively long to compute and we run smaller experiments (see Appendix F).”
BackgroundMethodInfluentialExplainable AI: Interpreting, Explaining and Visualizing Deep Learning
(2019) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-28954-6 · Cited by 1,377
Homer, N., Szelinger, S., Redman, M., Duggan, D., Tembe, W., Muehling, J., Pearson, J. V., Stephan, D. A., Nelson, S. F., & Craig, D. W. (2008) · PLoS Genetics · doi:10.1371/journal.pgen.1000167 · Cited by 1,221 · cited 1×
“(Homer et al. 2008) demonstrated a successful membership inference attack on aggregated genotype data provided by the US National Institutes of Health and other organizations.”
BackgroundPrivacy Risk in Machine Learning: Analyzing the Connection to Overfitting
Yeom, S., Giacomelli, I., Fredrikson, M., & Jha, S. (2018) · 2018 IEEE 31st Computer Security Foundations Symposium (CSF) · doi:10.1109/csf.2018.00027 · Cited by 1,001
Fooling LIME and SHAP: Adversarial Attacks on Post hoc Explanation Methods
Slack, D., Hilgard, S., Jia, E., Singh, S., & Lakkaraju, H. (2020) · Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society · doi:10.1145/3375627.3375830 · Cited by 900
Rationalizing Neural Predictions
Lei, T., Barzilay, R., & Jaakkola, T. (2016) · Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d16-1011 · Cited by 777
The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networks
Carlini, N., Liu, C., Erlingsson, Ú., Kos, J., & Song, D. (2018) · arXiv · doi:10.48550/arxiv.1802.08232 · Cited by 629 · cited 1×
“Our findings can be explained by earlier observations that training set outliers are likelier to be “memorized” and thus less generalized [11]; however, such memorization leaves minority populations vulnerable to privacy risks.”
ResultKaur, H., Nori, H., Jenkins, S., Caruana, R., Wallach, H., & Wortman Vaughan, J. (2020) · Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3313831.3376219 · Cited by 564 · cited 3×
“The layer sizes are [20, 20, 20, 20, 2].”
BackgroundInfluentialSemi-supervised Knowledge Transfer for Deep Learning from Private Training Data
Papernot, N., Abadi, M., Erlingsson, Ú., Goodfellow, I., & Talwar, K. (2016) · arXiv · doi:10.48550/arxiv.1610.05755 · Cited by 492
Machine Learning with Membership Privacy using Adversarial Regularization
Nasr, M., Shokri, R., & Houmansadr, A. (2018) · Proceedings of the 2018 ACM SIGSAC Conference on Computer and Communications Security · doi:10.1145/3243734.3243855 · Cited by 458 · cited 3×
“[29] M Nasr, R Shokri, and A Houmansadr.”
BackgroundMethodSelf-citationTesting the manifold hypothesis
Fefferman, C., Mitter, S., & Narayanan, H. (2016) · Journal of the American Mathematical Society · doi:10.1090/jams/852 · Cited by 430 · cited 2×
“This is due to an interesting connection between perturbationbased model explanations and the data-manifold hypothesis [17].”
BackgroundAuditing black-box models for indirect influence
Adler, P., Falk, C., Friedler, S. A., Nix, T., Rybeck, G., Scheidegger, C., Smith, B., & Venkatasubramanian, S. (2018) · Knowledge and Information Systems · doi:10.1007/s10115-017-1116-3 · Cited by 280
Auditing Black-Box Models for Indirect Influence
Adler, P., Falk, C., Friedler, S. A., Rybeck, G., Scheidegger, C., Smith, B., & Venkatasubramanian, S. (2016) · 2016 IEEE 16th International Conference on Data Mining (ICDM) · doi:10.1109/icdm.2016.0011 · Cited by 280
Exposed! A Survey of Attacks on Private Data
Dwork, C., Smith, A., Steinke, T., & Ullman, J. (2017) · Annual Review of Statistics and Its Application · doi:10.1146/annurev-statistics-060116-054123 · Cited by 259 · cited 2×
“For example-based explanations, our algorithms can reconstruct a large fraction of the training data (referred to as dataset reconstruction attacks [16]).”
BackgroundMethodPrivacy Risks of Securing Machine Learning Models against Adversarial Examples
Song, L., Shokri, R., & Mittal, P. (2019) · Proceedings of the 2019 ACM SIGSAC Conference on Computer and Communications Security · doi:10.1145/3319535.3354211 · Cited by 229 · cited 2×
“However, robustness results in vulnerability to membership inference [43].”
BackgroundSelf-citationProblems with Shapley-value-based explanations as feature importance measures
Kumar, I. E., Venkatasubramanian, S., Scheidegger, C., & Friedler, S. (2020) · arXiv · doi:10.48550/arxiv.2002.11097 · Cited by 228 · cited 2×
“[25] I E Kumar, S Venkatasubramanian, C Scheidegger, and S Friedler.”
BackgroundHow to Explain Individual Classification Decisions
(2010) · Journal of Machine Learning Research · doi:10.5555/1756006.1859912 · Cited by 54
Cited by
Towards Personalized Federated Learning
Tan, A. Z., Yu, H., Cui, L., & Yang, Q. (2023) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2022.3160699 · Cited by 1,182
Explainable machine learning in deployment
Bhatt, U., Xiang, A., Sharma, S., Weller, A., Taly, A., Jia, Y., Ghosh, J., Puri, R., Moura, J. M. F., & Eckersley, P. (2020) · Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency · doi:10.1145/3351095.3375624 · Cited by 633
Membership Inference Attacks on Machine Learning: A Survey
Hu, H., Salcic, Z., Sun, L., Dobbie, G., Yu, P. S., & Zhang, X. (2022) · ACM Computing Surveys · doi:10.1145/3523273 · Cited by 564
A Survey of Algorithmic Recourse: Contrastive Explanations and Consequential Recommendations
Karimi, A.-H., Barthe, G., Schölkopf, B., & Valera, I. (2023) · ACM Computing Surveys · doi:10.1145/3527848 · Cited by 301
Counterfactual Explanations and Algorithmic Recourses for Machine Learning: A Review
Verma, S., Boonsanong, V., Hoang, M., Hines, K. E., Dickerson, J. P., & Shah, C. (2020) · arXiv · doi:10.48550/arxiv.2010.10596 · Cited by 267
Metadata
- Authors
- Shokri, R., Strobel, M., & Zick, Y.
- Venue
- Proceedings of the 2021 AAAI/ACM Conference on AI, Ethics, and Society
- Publisher
- ACM
- Published
- 2021-07-21
- Pages
- 231-241
- ISSN
- 3065-8365 (ISSN-L)
- OpenAlex
- W3042265470
- Cited by
- 111
- References
- 28
- FWCI
- 8.76
- Type
- conference paper
- Language
- en
- Open access
- bronze
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.