Adversarial Machine Learning Applied to Intrusion and Malware Scenarios: A Systematic Review

Abstract

Cyber-security is the practice of protecting computing systems and networks from digital attacks, which are a rising concern in the Information Age. With the growing pace at which new attacks are developed, conventional signature based attack detection methods are often not enough, and machine learning poses as a potential solution. Adversarial machine learning is a research area that examines both the generation and detection of adversarial examples, which are inputs specially crafted to deceive classifiers, and has been extensively studied specifically in the area of image recognition, where minor modifications are performed on images that cause a classifier to produce incorrect predictions. However, in other fields, such as intrusion and malware detection, the exploration of such methods is still growing. The aim of this survey is to explore works that apply adversarial machine learning concepts to intrusion and malware detection scenarios. We concluded that a wide variety of attacks were tested and proven effective in malware and intrusion detection, although their practicality was not tested in intrusion scenarios. Adversarial defenses were substantially less explored, although their effectiveness was also proven at resisting adversarial attacks. We also concluded that, contrarily to malware scenarios, the variety of datasets in intrusion scenarios is still very small, with the most used dataset being greatly outdated.

Modeling Realistic Adversarial Attacks against Network Intrusion Detection Systems

“…ML-NIDS are gaining popularity [11,30], with the typical consequence that real attackers are turning their attention to the vulnerabilities of some ML components. This paper considers the so called adversarial attacks against ML-NIDS [62,92], and does not make any assumption on the specific ML-algorithm, nor on the analyzed data type of the ML-NIDS.…”

“…By leveraging such information it is possible to understand the decision boundaries of the ML model and to craft specific samples that thwart the detection mechanism. These attacks are also known as white-box attacks [62].…”

Tight Arms Race: Overview of Current Malware Threats and Trends in Their Detection

“…Lastly, evasion has also been a major topic among the researchers and practitioners conjugating ML with cybersecurity. As an example, the survey [31] focused on the malware evading ML-based detection techniques. It explored the adversarial attacks and the applications of adversarial ML for malware detection.…”

“…• limited interest in information hiding: as shown, only one recent work dealt with information hiding (specifically, in the context of mobile devices). As modern mal- [23] x x x x x [16] x x x x x [24] x x [22] x x fileless [31] x adversarial ML [17] x tools [25] x AI [14] x x x evasion, tools [34] x x x evasion [18] x x evasion [21] x x APT [28] x x cybersecurity [19] x evasion [26] x x x x [27] x x cybersecurity [36] x x x visualisation [30] x x [11] x x behavior analysis, visualisation [12] x analysis [37] x x [33] x x x evasion [29] x x x [15] x x x x x [20] x x stealth malware [32] x x x C&C communication [35] x x x OS openness [38] x x [39] x visualisation ware is increasingly exploiting some form of steganography, information hiding and obfuscation to launch attacks or exfiltrate data [42], [43], this consolidated trend should be taken into account. • lack of sufficient coverage of new threats: despite the vivacity of the topic, many works continue to focus on the "legacy" hazards, e.g., phishing.…”

Artificial Intelligence Security Threat, Crime, and Forensics: Taxonomy and Open Issues

“…self-driving car) by proposing a possible attack to misclassify signs on the road. Several studies also have represented that AEs can be generated and utilized for disturbing malware detection [95], [96] and intrusion detection [97], [98]; the forensic investigators should be aware of data injection crime. If perpetrators have permission to modify or delete some training data, they can perform fatal attacks to AI system; it is the data modification crime.…”