What’s Needed to Prevent Cyberbiosecurity Threats and Protect Vulnerable Countries
By Eleonore Pauwels
The CT lung scan showed the ravaging signs of COVID-19 and the inflammatory response, the cytokine storm. But what if the CT scan was wrong?
This is no passing concern. In 2018, a malicious attack was designed to hack hospital CT scans, generating false lung tumours that conformed to a patient’s unique anatomy, leading to a misdiagnosis rate in excess of 90%. Furthermore, researchers at Harvard University tested adversarial attacks against algorithms used to diagnose skin cancer images, demonstrating that such attacks required only modifying a few pixels in the original biopsy picture to corrupt a diagnosis.
These examples are just a sampling of how AI can automate the manipulation of medical datasets, expanding a cyberattack’s impact through health and biotech industries. Those attacks exist in an era of hybrid security risks where pandemic threats converge with the weaponization of powerful dual-use technologies. Developed for beneficial purposes, dual-use technologies can also inherently cause harm, either accidental “unintended harms,” or as a result of deliberate malicious intent. The convergence of cyber- and biosecurity threats reduce our ability, in a global health crisis, to trust what the digital pictures of our bodies tell us about a complex virus like COVID-19 and how fast we analyse pathogens in future crises, natural or man-made.
Photo courtesy of Keoni Cabral.