The quantity of medical imaging data has exploded over recent years, becoming an integral part of healthcare. But securing this sensitive information is a problem that continues to plague organizations.
Many have turned to digitally watermarking medical images in an effort to protect their integrity, prove ownership and prevent unauthorized access. And a handful of computer scientists from the Czech Republic recently performed a deep dive review of a popular method known as "zero" watermarking.
“The main advantage of the zero watermarking method in medical imaging is that the watermark is not embedded in the image itself. So, unlike traditional watermarking techniques, the watermarking process does not apply any modification to the image, thus avoiding any distortion of the image,” Aleš Roček, with Masaryk University’s Institute of Computer Science, explained Jan. 7 in the Journal of Digital Imaging.
From there, hidden features are extracted from the host image and combined with watermark information, such as a logo. That data is encrypted and used to produce a key.
The group tested if zero watermarking could hold up during three simulated attacks: modifying a secured image, replacing a secured image with neighboring CT images, and replacing an image with those of the same body part, but a different patient.
Roček and colleagues found that the technique is “suitable” for checking the integrity of digital images, but falls short for securing them.
“Zero watermarking is not able to reliably … distinguish a modified image belonging to the author from somebody else’s similar image,” the authors explained. “Especially in the medical field, it is not capable of distinguishing images of the same part of the body originating from two different patients.”
Zero watermarking does, however, excel at reflecting the degree to which an image has been modified. With this in mind, the researchers said it can likely be used as a “measure of image similarity,” but will require further testing.