%0 Journal Article %1 moisejevs2019adversarial %A Moisejevs, Ilja %D 2019 %E Editor, %J International Journal of Artificial Intelligence and Expert Systems (IJAE) %K Adversarial Attacks Attacks, Detection Evasion Intrusion Learning, Machine Poisoning Systems, %N 3 %P 44 - 62 %T Adversarial Attacks and Defenses in Intrusion Detection Systems: A Survey %U http://www.cscjournals.org/library/manuscriptinfo.php?mc=IJAE-186 %V 8 %X The world is becoming more digitized and inter-connected by the day and securing our digital infrastructure is not a topic we can take lightly anymore. Intrusion detection systems (IDSs) have been an integral part of the cybersecurity stack ever since their introduction in the 1980s. Traditionally such systems have relied on signatures and heuristics, however, recently growing demand for scalability, advances in computational power, and increasing dataset availability, have paved the way for machine learning approaches. The challenge is that even though machine learning can do a better job at detecting intrusions in normal conditions - it itself is left vulnerable to adaptive adversaries who understand how these systems work and "think". In this survey we review the different kinds of attacks such an adversary can mount on IDSs, and perhaps more importantly, the various defenses available for making IDSs more robust. We start by proving some historic context on the matter and introducing the basic taxonomy of adversarial machine learning, before diving into the methods, attacks and defenses in the second part of the write-up.

@article{moisejevs2019adversarial, abstract = {The world is becoming more digitized and inter-connected by the day and securing our digital infrastructure is not a topic we can take lightly anymore. Intrusion detection systems (IDSs) have been an integral part of the cybersecurity stack ever since their introduction in the 1980s. Traditionally such systems have relied on signatures and heuristics, however, recently growing demand for scalability, advances in computational power, and increasing dataset availability, have paved the way for machine learning approaches. The challenge is that even though machine learning can do a better job at detecting intrusions in normal conditions - it itself is left vulnerable to adaptive adversaries who understand how these systems work and "think". In this survey we review the different kinds of attacks such an adversary can mount on IDSs, and perhaps more importantly, the various defenses available for making IDSs more robust. We start by proving some historic context on the matter and introducing the basic taxonomy of adversarial machine learning, before diving into the methods, attacks and defenses in the second part of the write-up.}, added-at = {2020-02-03T07:04:21.000+0100}, author = {Moisejevs, Ilja}, biburl = {https://www.bibsonomy.org/bibtex/2d0a8023ddb6e7136546e9d1c23a764db/cscjournals}, editor = {Editor}, interhash = {3c193af6f4fc3a419db651130bb48f25}, intrahash = {d0a8023ddb6e7136546e9d1c23a764db}, issn = {2180-124X}, journal = {International Journal of Artificial Intelligence and Expert Systems (IJAE)}, keywords = {Adversarial Attacks Attacks, Detection Evasion Intrusion Learning, Machine Poisoning Systems,}, language = {English}, month = {October}, number = 3, pages = {44 - 62}, timestamp = {2020-02-03T07:04:21.000+0100}, title = {Adversarial Attacks and Defenses in Intrusion Detection Systems: A Survey}, url = {http://www.cscjournals.org/library/manuscriptinfo.php?mc=IJAE-186}, volume = 8, year = 2019 }