Pattern classification under attack on spam filtering
Authors:
- 1 November 2015
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 197-201
Abstract
Spam Filtering is an adversary application in which data can be purposely employed by humans to attenuate their operation. Statistical spam filters are manifest to be vulnerable to adversarial attacks. To evaluate security issues related to spam filtering numerous machine learning systems are used. For adversary applications some Pattern classification systems are ordinarily used, since these systems are based on classical theory and design approaches do not take into account adversarial settings. Pattern classification system display vulnerabilities (i.e. a weakness that grants an attacker to reduce assurance on system's information) to several potential attacks, allowing adversaries to attenuate their effectiveness. In this paper, security evaluation of spam email using pattern classifier during an attack is addressed which degrade the performance of the system. Additionally a model of the adversary is used that allows defining spam attack scenario.Keywords
This publication has 9 references indexed in Scilit:
- Security Evaluation of Pattern Classifiers under AttackIEEE Transactions on Knowledge and Data Engineering, 2014
- Adversarial machine learning2011
- Multimodal fusion vulnerability to non-zero effort (spoof) imposters2010
- The security of machine learningMachine Learning, 2010
- A framework for quantitative security analysis of machine learning2009
- Can machine learning be secure?2006
- A framework for the evaluation of intrusion detection systems2006
- Adversarial learning2005
- Adversarial classification2004