Abstract
Deep learning methods are usually trained via a gradient-descent based procedure, which can be efficient as it is not only end-to-end but also suitable for large quantities of data. However, gradient-based learning is vulnerable to adversarial attacks – which account for unperceivable changes in the input data to misguide a trained model. Though a plethora of work explored the adversarial learning (attacks and defences) in image datasets, the exploration of adversarial learning in tabular datasets has seen little attention. In this work, we study adversarial learning in tabular datasets. We investigate the role of discretization and demonstrate that discretizing numeric attributes offers a strong defence mechanism. The main contribution of this work is the proposition of two new defence algorithms for numeric tabular datasets, that utilize cut-points obtained from discretization, to forge a defence against various forms of adversarial attacks. We evaluate the effectiveness of our proposed method on a wide range of machine learning datasets and demonstrate that the proposed algorithms lead to a state-of-the-art defence strategy on tabular datasets.
J. Zhou and N. Zaidi—Equal Contribution.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
This is one reason, why adversarial attacks against tabular data are not prevalent as compared to against image datasets.
- 2.
Note, we have combined the two proposed algorithms into one due to space constraints.
- 3.
The effectiveness of adversarial training loss component is also studied in ablation study in Sect. 4.4.
References
Ballet, V., Renard, X., Aigrain, J., Laugel, T., Frossard, P., Detyniecki, M.: Imperceptible adversarial attacks on tabular data. arXiv preprint arXiv:1911.03274 (2019)
Buckman, J., Roy, A., Raffel, C., Goodfellow, I.: Thermometer encoding: one hot way to resist adversarial examples. In: International Conference on Learning Representations (2018)
Goodfellow, I.J., Shlens, J., Szegedy, C.: Explaining and harnessing adversarial examples. arXiv preprint arXiv:1412.6572 (2014)
Kurakin, A., Goodfellow, I., Bengio, S.: Adversarial machine learning at scale. arXiv preprint arXiv:1611.01236 (2016)
Madry, A., Makelov, A., Schmidt, L., Tsipras, D., Vladu, A.: Towards deep learning models resistant to adversarial attacks. arXiv preprint arXiv:1706.06083 (2017)
Moosavi-Dezfooli, S.M., Fawzi, A., Frossard, P.: DeepFool: a simple and accurate method to fool deep neural networks. In: CVPR (2016)
Qiu, S., Liu, Q., Zhou, S., Wu, C.: Review of artificial intelligence adversarial attack and defense technologies. Appl. Sci. 9(5), 909 (2019)
Sulewski, P.: Equal-bin-width histogram versus equal-bin-count histogram. J. Appl. Stat. 48(12), 2092–2111 (2021)
Yang, S., Guo, T., Wang, Y., Xu, C.: Adversarial robustness through disentangled representations. In: Proceedings of the AAAI Conference on Artificial Intelligence, vol. 35, pp. 3145–3153 (2021)
Yang, Y., Webb, G.I.: Discretization for Naive-Bayes learning: managing discretization bias and variance. Mach. Learn. 74(1), 39–74 (2009)
Zaidi, N.A., Du, Y., Webb, G.I.: On the effectiveness of discretizing quantitative attributes in linear classifiers. IEEE Access 8, 198856–198871 (2020). https://doi.org/10.1109/ACCESS.2020.3034955
Zhang, H., Yu, Y., Jiao, J., Xing, E., El Ghaoui, L., Jordan, M.: Theoretically principled trade-off between robustness and accuracy. In: ICML (2019)
Zhang, Y., Zaidi, N.A., Zhou, J., Li, G.: GANBLR: a tabular data generation model. In: 2021 IEEE International Conference on Data Mining (ICDM), pp. 181–190. IEEE (2021)
Zhang, Y., Zaidi, N.A., Zhou, J., Li, G.: GANBLR++: incorporating capacity to generate numeric attributes and leveraging unrestricted Bayesian networks. In: Proceedings of the 2022 SIAM International Conference on Data Mining (2022)
Zhou, M., Wu, J., Liu, Y., Liu, S., Zhu, C.: DAST: data-free substitute training for adversarial attacks. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 234–243 (2020)
Acknowledgement
This research is partially supported by the National Natural Science Fund of China (Project No. 71871090).
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Zhou, J., Zaidi, N., Zhang, Y., Li, G. (2022). Discretization Inspired Defence Algorithm Against Adversarial Attacks on Tabular Data. In: Gama, J., Li, T., Yu, Y., Chen, E., Zheng, Y., Teng, F. (eds) Advances in Knowledge Discovery and Data Mining. PAKDD 2022. Lecture Notes in Computer Science(), vol 13281. Springer, Cham. https://doi.org/10.1007/978-3-031-05936-0_29
Download citation
DOI: https://doi.org/10.1007/978-3-031-05936-0_29
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-05935-3
Online ISBN: 978-3-031-05936-0
eBook Packages: Computer ScienceComputer Science (R0)