Journal article
Query strategies for evading convex-inducing classifiers
B Nelson, BIP Rubinstein, L Huang, AD Joseph, SJ Lee, S Rao, JD Tygar
Journal of Machine Learning Research | MICROTOME PUBL | Published : 2012
Abstract
Classifiers are often used to detect miscreant activities. We study how an adversary can systematically query a classifier to elicit information that allows the attacker to evade detection while incurring a near-minimal cost of modifying their intended malfeasance. We generalize the theory of Lowd and Meek (2005) to the family of convex-inducing classifiers that partition their feature space into two sets, one of which is convex. We present query algorithms for this family that construct undetected instances of approximately minimal cost using only polynomially-many queries in the dimension of the space and in the level of approximation. Our results demonstrate that nearoptimal evasion can b..
View full abstractGrants
Awarded by National Science Foundation (NSF)
Awarded by AFOSR
Awarded by California state MICRO
Awarded by DHS HSARPA
Awarded by NSF award
Funding Acknowledgements
We gratefully acknowledge the support of our sponsors. This work was supported in part by TRUST (Team for Research in Ubiquitous Secure Technology), which receives support from the National Science Foundation (NSF award #CCF-0424422) and AFOSR (#FA9550-06-1-0244); RAD Lab, which receives support from California state MICRO grants (#06-148 and #07-012); DETER-lab (cyber-DEfense Technology Experimental Research laboratory), which receives support from DHS HSARPA (#022412) and AFOSR (#FA9550-07-1-0501); NSF award #DMS-0707060; the Siebel Scholars Foundation; and the following organizations: Amazon, BT, Cisco, DoCoMo USA Labs, EADS, ESCHER, Facebook, Google, HP, IBM, iCAST, Intel, Microsoft, NetApp, ORNL, Pirelli, Qualcomm, Sun, Symantec, TCS, Telecom Italia, United Technologies, and VMware. The opinions expressed in this paper are solely those of the authors and do not necessarily reflect the opinions of any funding agency, the State of California, or the U.S. government.