Dongguk University Develops Defense System for Medical Digital Twins

Dongguk University has developed a new defense system called Wavelet-Based Adversarial Training (WBAD) to protect medical digital twins from cyberattacks, announced in a press release. This innovative method enables medical digital twins to achieve 98% accuracy in breast cancer prediction, even under adversarial attacks.

Medical digital twins are virtual models of the human body used to predict diseases with high accuracy. However, they are vulnerable to cyberattacks that can manipulate data and lead to incorrect diagnoses. To address this, researchers from Dongguk University, in collaboration with Oregon State University, developed WBAD, which combines wavelet denoising and adversarial training to enhance the security and reliability of these digital models.

The WBAD system was tested on a breast cancer diagnostic model using thermography images. Initially, the model achieved 92% accuracy, but its accuracy dropped to 5% when subjected to adversarial attacks. By implementing a two-layer defense mechanism, including wavelet denoising and adversarial training, the model's accuracy was restored to 98% against Fast Gradient Sign Method attacks, 93% against Projected Gradient Descent attacks, and 90% against Carlini & Wagner attacks.

This breakthrough provides a comprehensive and effective defense against cyberattacks, ensuring safer and more reliable medical digital twins for healthcare applications.