Adversarial examples: A survey of attacks and defenses in deep learning-enabled cybersecurity systems

• Published in: Expert systems with applications Vol. 238; p. 122223
• Main Authors: Macas, Mayra, Wu, Chunming, Fuertes, Walter
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 15-03-2024
• Subjects: Adversarial examples; Adversarial machine learning; Cyber threats; Cybersecurity; Deep learning; Deep learning; Adversarial examples; Cybersecurity; Adversarial machine learning; Cyber threats
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2023.122223

Over the last few years, the adoption of machine learning in a wide range of domains has been remarkable. Deep learning, in particular, has been extensively used to drive applications and services in specializations such as computer vision, natural language processing, machine translation, and cybersecurity, producing results that are comparable to or even surpass the performance of human experts. Nevertheless, machine learning systems are vulnerable to adversarial attacks, especially in nonstationary environments where actual adversaries exist, such as the cybersecurity domain. In this work, we comprehensively survey and present the latest research on attacks based on adversarial examples against deep learning-based cybersecurity systems, highlighting the risks they pose and promoting efficient countermeasures. To that end, adversarial attack methods are first categorized according to where they occur and the attacker’s goals and capabilities. Then, specific attacks based on adversarial examples and the respective defensive methods are reviewed in detail within the framework of eight principal cybersecurity application categories. Finally, the main trends in recent research are outlined, and the impact of recent advancements in adversarial machine learning is explored to provide guidelines and directions for future research in cybersecurity. In summary, this work is the first to systematically analyze adversarial example-based attacks in the cybersecurity field, discuss possible defenses, and highlight promising directions for future research.
•A taxonomy of cybersecurity applications is established.•Adversarial machine learning is systematically overviewed.•An extensive, curated list of cybersecurity-related datasets is provided.•Methods for generating adversarial examples and suitable defenses are reviewed.•Challenges, open issues, and future research directions are outlined.