Scalable Optimization of Randomized Operational Decisions in Adversarial Classification Settings

Bo Li, Yevgeniy Vorobeychik
Proceedings of the Eighteenth International Conference on Artificial Intelligence and Statistics, PMLR 38:599-607, 2015.

Abstract

When learning, such as classification, is used in adversarial settings, such as intrusion detection, intelligent adversaries will attempt to evade the resulting policies. The literature on adversarial machine learning aims to develop learning algorithms which are robust to such adversarial evasion, but exhibits two significant limitations: a) failure to account for operational constraints and b) a restriction that decisions are deterministic. To overcome these limitations, we introduce a conceptual separation between learning, used to infer attacker preferences, and operational decisions, which account for adversarial evasion, enforce operational constraints, and naturally admit randomization. Our approach gives rise to an intractably large linear program. To overcome scalability limitations, we introduce a novel method for estimating a compact parity basis representation for the operational decision function. Additionally, we develop an iterative constraint generation approach which embeds adversary’s best response calculation, to arrive at a scalable algorithm for computing near-optimal randomized operational decisions. Extensive experiments demonstrate the efficacy of our approach.

Cite this Paper

BibTeX
@InProceedings{pmlr-v38-li15a, title = {{Scalable Optimization of Randomized Operational Decisions in Adversarial Classification Settings}}, author = {Li, Bo and Vorobeychik, Yevgeniy}, booktitle = {Proceedings of the Eighteenth International Conference on Artificial Intelligence and Statistics}, pages = {599--607}, year = {2015}, editor = {Lebanon, Guy and Vishwanathan, S. V. N.}, volume = {38}, series = {Proceedings of Machine Learning Research}, address = {San Diego, California, USA}, month = {09--12 May}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v38/li15a.pdf}, url = {https://proceedings.mlr.press/v38/li15a.html}, abstract = {When learning, such as classification, is used in adversarial settings, such as intrusion detection, intelligent adversaries will attempt to evade the resulting policies. The literature on adversarial machine learning aims to develop learning algorithms which are robust to such adversarial evasion, but exhibits two significant limitations: a) failure to account for operational constraints and b) a restriction that decisions are deterministic. To overcome these limitations, we introduce a conceptual separation between learning, used to infer attacker preferences, and operational decisions, which account for adversarial evasion, enforce operational constraints, and naturally admit randomization. Our approach gives rise to an intractably large linear program. To overcome scalability limitations, we introduce a novel method for estimating a compact parity basis representation for the operational decision function. Additionally, we develop an iterative constraint generation approach which embeds adversary’s best response calculation, to arrive at a scalable algorithm for computing near-optimal randomized operational decisions. Extensive experiments demonstrate the efficacy of our approach.} }
Endnote
%0 Conference Paper %T Scalable Optimization of Randomized Operational Decisions in Adversarial Classification Settings %A Bo Li %A Yevgeniy Vorobeychik %B Proceedings of the Eighteenth International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2015 %E Guy Lebanon %E S. V. N. Vishwanathan %F pmlr-v38-li15a %I PMLR %P 599--607 %U https://proceedings.mlr.press/v38/li15a.html %V 38 %X When learning, such as classification, is used in adversarial settings, such as intrusion detection, intelligent adversaries will attempt to evade the resulting policies. The literature on adversarial machine learning aims to develop learning algorithms which are robust to such adversarial evasion, but exhibits two significant limitations: a) failure to account for operational constraints and b) a restriction that decisions are deterministic. To overcome these limitations, we introduce a conceptual separation between learning, used to infer attacker preferences, and operational decisions, which account for adversarial evasion, enforce operational constraints, and naturally admit randomization. Our approach gives rise to an intractably large linear program. To overcome scalability limitations, we introduce a novel method for estimating a compact parity basis representation for the operational decision function. Additionally, we develop an iterative constraint generation approach which embeds adversary’s best response calculation, to arrive at a scalable algorithm for computing near-optimal randomized operational decisions. Extensive experiments demonstrate the efficacy of our approach.
RIS
TY - CPAPER TI - Scalable Optimization of Randomized Operational Decisions in Adversarial Classification Settings AU - Bo Li AU - Yevgeniy Vorobeychik BT - Proceedings of the Eighteenth International Conference on Artificial Intelligence and Statistics DA - 2015/02/21 ED - Guy Lebanon ED - S. V. N. Vishwanathan ID - pmlr-v38-li15a PB - PMLR DP - Proceedings of Machine Learning Research VL - 38 SP - 599 EP - 607 L1 - http://proceedings.mlr.press/v38/li15a.pdf UR - https://proceedings.mlr.press/v38/li15a.html AB - When learning, such as classification, is used in adversarial settings, such as intrusion detection, intelligent adversaries will attempt to evade the resulting policies. The literature on adversarial machine learning aims to develop learning algorithms which are robust to such adversarial evasion, but exhibits two significant limitations: a) failure to account for operational constraints and b) a restriction that decisions are deterministic. To overcome these limitations, we introduce a conceptual separation between learning, used to infer attacker preferences, and operational decisions, which account for adversarial evasion, enforce operational constraints, and naturally admit randomization. Our approach gives rise to an intractably large linear program. To overcome scalability limitations, we introduce a novel method for estimating a compact parity basis representation for the operational decision function. Additionally, we develop an iterative constraint generation approach which embeds adversary’s best response calculation, to arrive at a scalable algorithm for computing near-optimal randomized operational decisions. Extensive experiments demonstrate the efficacy of our approach. ER -
APA
Li, B. & Vorobeychik, Y.. (2015). Scalable Optimization of Randomized Operational Decisions in Adversarial Classification Settings. Proceedings of the Eighteenth International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 38:599-607 Available from https://proceedings.mlr.press/v38/li15a.html.

Related Material