Adversarial Robustness Guarantees for Classification with Gaussian Processes

Arno Blaas, Andrea Patane, Luca Laurenti, Luca Cardelli, Marta Kwiatkowska, Stephen Roberts
Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, PMLR 108:3372-3382, 2020.

Abstract

We investigate adversarial robustness of Gaussian Process classification (GPC) models. Specifically, given a compact subset of the input space $T\subseteq \mathbb{R}^d$ enclosing a test point $x^*$ and a GPC trained on a dataset $\mathcal{D}$, we aim to compute the minimum and the maximum classification probability for the GPC over all the points in $T$.In order to do so, we show how functions lower- and upper-bounding the GPC output in $T$ can be derived, and implement those in a branch and bound optimisation algorithm. For any error threshold $\epsilon > 0$ selected \emph{a priori}, we show that our algorithm is guaranteed to reach values $\epsilon$-close to the actual values in finitely many iterations.We apply our method to investigate the robustness of GPC models on a 2D synthetic dataset, the SPAM dataset and a subset of the MNIST dataset, providing comparisons of different GPC training techniques, and show how our method can be used for interpretability analysis. Our empirical analysis suggests that GPC robustness increases with more accurate posterior estimation.

Cite this Paper


BibTeX
@InProceedings{pmlr-v108-blaas20a, title = {Adversarial Robustness Guarantees for Classification with Gaussian Processes}, author = {Blaas, Arno and Patane, Andrea and Laurenti, Luca and Cardelli, Luca and Kwiatkowska, Marta and Roberts, Stephen}, booktitle = {Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics}, pages = {3372--3382}, year = {2020}, editor = {Silvia Chiappa and Roberto Calandra}, volume = {108}, series = {Proceedings of Machine Learning Research}, month = {26--28 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v108/blaas20a/blaas20a.pdf}, url = { http://proceedings.mlr.press/v108/blaas20a.html }, abstract = {We investigate adversarial robustness of Gaussian Process classification (GPC) models. Specifically, given a compact subset of the input space $T\subseteq \mathbb{R}^d$ enclosing a test point $x^*$ and a GPC trained on a dataset $\mathcal{D}$, we aim to compute the minimum and the maximum classification probability for the GPC over all the points in $T$.In order to do so, we show how functions lower- and upper-bounding the GPC output in $T$ can be derived, and implement those in a branch and bound optimisation algorithm. For any error threshold $\epsilon > 0$ selected \emph{a priori}, we show that our algorithm is guaranteed to reach values $\epsilon$-close to the actual values in finitely many iterations.We apply our method to investigate the robustness of GPC models on a 2D synthetic dataset, the SPAM dataset and a subset of the MNIST dataset, providing comparisons of different GPC training techniques, and show how our method can be used for interpretability analysis. Our empirical analysis suggests that GPC robustness increases with more accurate posterior estimation.} }
Endnote
%0 Conference Paper %T Adversarial Robustness Guarantees for Classification with Gaussian Processes %A Arno Blaas %A Andrea Patane %A Luca Laurenti %A Luca Cardelli %A Marta Kwiatkowska %A Stephen Roberts %B Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2020 %E Silvia Chiappa %E Roberto Calandra %F pmlr-v108-blaas20a %I PMLR %P 3372--3382 %U http://proceedings.mlr.press/v108/blaas20a.html %V 108 %X We investigate adversarial robustness of Gaussian Process classification (GPC) models. Specifically, given a compact subset of the input space $T\subseteq \mathbb{R}^d$ enclosing a test point $x^*$ and a GPC trained on a dataset $\mathcal{D}$, we aim to compute the minimum and the maximum classification probability for the GPC over all the points in $T$.In order to do so, we show how functions lower- and upper-bounding the GPC output in $T$ can be derived, and implement those in a branch and bound optimisation algorithm. For any error threshold $\epsilon > 0$ selected \emph{a priori}, we show that our algorithm is guaranteed to reach values $\epsilon$-close to the actual values in finitely many iterations.We apply our method to investigate the robustness of GPC models on a 2D synthetic dataset, the SPAM dataset and a subset of the MNIST dataset, providing comparisons of different GPC training techniques, and show how our method can be used for interpretability analysis. Our empirical analysis suggests that GPC robustness increases with more accurate posterior estimation.
APA
Blaas, A., Patane, A., Laurenti, L., Cardelli, L., Kwiatkowska, M. & Roberts, S.. (2020). Adversarial Robustness Guarantees for Classification with Gaussian Processes. Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 108:3372-3382 Available from http://proceedings.mlr.press/v108/blaas20a.html .

Related Material