Adversarial Edit Attacks for Tree Data

Paaßen B (2019)
In: Proceedings of the 20th International Conference on Intelligent Data Engineering and Automated Learning (IDEAL 2019). Yin H, Camacho D, Tino P (Eds); Lecture Notes in Computer Science, 11871. Cham: Springer: 359-366.

Konferenzbeitrag | Veröffentlicht | Englisch
 
Herausgeber*in
Yin, Hujun; Camacho, David; Tino, Peter
Abstract / Bemerkung
Many machine learning models can be attacked with adversarial examples, i.e. inputs close to correctly classified examples that are classified incorrectly. However, most research on adversarial attacks to date is limited to vectorial data, in particular image data. In this contribution, we extend the field by introducing adversarial edit attacks for tree-structured data with potential applications in medicine and automated program analysis. Our approach solely relies on the tree edit distance and a logarithmic number of black-box queries to the attacked classifier without any need for gradient information. We evaluate our approach on two programming and two biomedical data sets and show that many established tree classifiers, like tree-kernel-SVMs and recursive neural networks, can be attacked effectively.
Stichworte
Adversarial attacks; Tree edit distance; Structured data; Tree kernels; Recursive neural networks; Tree echo state networks
Erscheinungsjahr
2019
Titel des Konferenzbandes
Proceedings of the 20th International Conference on Intelligent Data Engineering and Automated Learning (IDEAL 2019)
Band
11871
Seite(n)
359-366
Konferenz
20th International Conference on Intelligent Data Engineering and Automated Learning (IDEAL 2019)
Konferenzort
Manchester, UK
Konferenzdatum
2019-11-14 – 2019-11-16
eISBN
978-3-030-33617-2
Page URI
https://pub.uni-bielefeld.de/record/2937053

Zitieren

Paaßen B. Adversarial Edit Attacks for Tree Data. In: Yin H, Camacho D, Tino P, eds. Proceedings of the 20th International Conference on Intelligent Data Engineering and Automated Learning (IDEAL 2019). Lecture Notes in Computer Science. Vol 11871. Cham: Springer; 2019: 359-366.
Paaßen, B. (2019). Adversarial Edit Attacks for Tree Data. In H. Yin, D. Camacho, & P. Tino (Eds.), Lecture Notes in Computer Science: Vol. 11871. Proceedings of the 20th International Conference on Intelligent Data Engineering and Automated Learning (IDEAL 2019) (pp. 359-366). Cham: Springer. doi:10.1007/978-3-030-33607-3_39
Paaßen, B. (2019). “Adversarial Edit Attacks for Tree Data” in Proceedings of the 20th International Conference on Intelligent Data Engineering and Automated Learning (IDEAL 2019), Yin, H., Camacho, D., and Tino, P. eds. Lecture Notes in Computer Science, vol. 11871, (Cham: Springer), 359-366.
Paaßen, B., 2019. Adversarial Edit Attacks for Tree Data. In H. Yin, D. Camacho, & P. Tino, eds. Proceedings of the 20th International Conference on Intelligent Data Engineering and Automated Learning (IDEAL 2019). Lecture Notes in Computer Science. no.11871 Cham: Springer, pp. 359-366.
B. Paaßen, “Adversarial Edit Attacks for Tree Data”, Proceedings of the 20th International Conference on Intelligent Data Engineering and Automated Learning (IDEAL 2019), H. Yin, D. Camacho, and P. Tino, eds., Lecture Notes in Computer Science, vol. 11871, Cham: Springer, 2019, pp.359-366.
Paaßen, B.: Adversarial Edit Attacks for Tree Data. In: Yin, H., Camacho, D., and Tino, P. (eds.) Proceedings of the 20th International Conference on Intelligent Data Engineering and Automated Learning (IDEAL 2019). Lecture Notes in Computer Science. 11871, p. 359-366. Springer, Cham (2019).
Paaßen, Benjamin. “Adversarial Edit Attacks for Tree Data”. Proceedings of the 20th International Conference on Intelligent Data Engineering and Automated Learning (IDEAL 2019). Ed. Hujun Yin, David Camacho, and Peter Tino. Cham: Springer, 2019.Vol. 11871. Lecture Notes in Computer Science. 359-366.
Link(s) zu Volltext(en)
Access Level
OA Open Access
Software:
Beschreibung
Software for edit distance computations

Export

Markieren/ Markierung löschen
Markierte Publikationen

Open Data PUB

Quellen

arXiv: 1908.09364

Suchen in

Google Scholar