To search, Click below search items. |
| All Published Papers Search Service |
|
|
|
Title |
Adversarial Threats in Healthcare: A Comprehensive Analysis of Vulnerabilities, Defense Mechanisms, and Recent Research |
|
Author |
Atrab A. Abd El-Aziz, Reda A El-Khoribi, Nour Eldeen Khalifa |
| Citation |
Vol. 26 No. 4 pp. 140-162 |
|
Abstract |
Recently, medical adversarial attacks had grave repercussions for both patients and healthcare providers. The resistance of deep learning-based medical models to adversarial attacks in dynamic situations is a topic of great interest since they are easily fooled using simple imperceptible perturbations called adversarial attacks. These attacks can exploit vulnerabilities in medical systems and can take various forms such as manipulating medical images, Medical Question Answering (MQA), or compromising electronic health records. These attacks resulted from some limitations in medical images such as the scarcity of high-quality imaging datasets and image labeling. Medical systems are very challenging due to their safety and accuracy are vital considerations. Although several defense techniques have been introduced to defend against these attacks, there is still a significant demand for more robust defense solutions. Notably, the domain of MQA and medical text is significantly under-researched, with virtually no established defense methods to counter adversarial threats in these areas. This paper contributes to the field by bridging the gap between adversarial attacks in medical images and medical text special MQA. It ensures that the review offers valuable insights and practical recommendations for researchers and practitioners in healthcare and cybersecurity. This paper covers various aspects of medical adversarial attack techniques, sources of vulnerabilities, defense strategies, and provides insights into the motivations and tactics of medical attacks. Finally, we provide recommendations for enhancing the resistance of medical systems to adversarial attacks. This review serves as a valuable resource for a diverse audience includes researchers and practitioners for healthcare and cybersecurity providers. It contributes to the collective knowledge base and offers practical recommendations for safeguarding medical systems against adversarial threats, particularly in the understudied areas of MQA and medical text. |
|
Keywords |
Adversarial Attacks, Deep Learning, Defense Methods, Medical Images, Medical Question Answering (MQA), Security. |
|
URL |
