Even before COVID-19 disrupted operations, organizations had accelerated their digital transformation initiatives to meet changing customer expectations. One sector that particularly embraced this shift is the healthcare sector, as organizations rapidly developed and adopted a range of digital health solutions, such as electronic health records and using artificial intelligence (AI) to aid drug discovery.
Healthcare is “an industry that had been moving forward with digitization under numerous different names and approaches well before the onset of COVID,” says Guy Becker, director of healthcare products management at cybersecurity company Sasa Software. However, this rapid digitization has also resulted in a sharp spike in criminal cyberattacks on the healthcare industry.
Check Point reported a global increase in attacks on organizations between November and December 2020. The report showed a 137% increase in East Asia, a 112% rise in Latin America, 67% in Europe, and 37% in North American healthcare organizations. In recent years, there has been a dramatic increase in cybersecurity incidents in the healthcare sector, such as computer virus infections, ransomware, and the theft and publication of patient data.
The reality is grimmer today, especially when you consider that scanned medical documents and other healthcare images often contain sensitive data. NTT Research recently held a hackathon to find ways to use attribute-based encryption (ABE) to address that situation and others.
“Metadata stored within medical images, including X-rays and CT scans, can disclose confidential information, like patient names, photographed body parts, and the medical centers or physicians involved, leading to patient identification,” explains Jean-Philippe Cabay, data scientist at NTT Global in Belgium, whose team won the hackathon. “Attribute-based encryption ensures that only authorized users with the appropriate attributes can access medical images, keeping them secure and private.”
Health Imaging Data Is a Hacker’s Goldmine
Hospitals and healthcare organizations are working to protect digital imaging and communications in medicine (DICOM) files, according to Becker. This development is a result of the convergence of several factors, increased attacks on healthcare due to its high value (worth at least 10 times more than credit card data on the Dark Web) and traditionally weak security posture, demand for heightened healthcare security by governments and the EU, increased need for remote healthcare services due to COVID, and a general digital transformation trend to streamline and digitize services.
In addition, the vulnerability presented by potentially malicious imaging files is enhanced by the growing risk of breached medical devices. For example, imaging machines operating within the hospital network can be compromised without the knowledge of the technicians and engineers looking after them. Such compromise could lead to malicious code being injected into clinical data and spread across a hospital’s network. Because imaging clinics and medical centers often need to transfer imaging data, a breach of such transactions could expose sensitive patient data, with devastating consequences.
Becker says the protection of sensitive imaging networks begins with the standard recommended measures: network segmentation, timely backups, frequent updating of systems and applications, the use of advanced intrusion detection and prevention systems, and regular employee education and training.
Some of these measures pose particular challenges for healthcare organizations. Healthcare systems have to be online 24/7, which makes frequent updating — and rebooting, or taking machines offline — an impossible requirement to meet. Chronic understaffing, which frequently reduces staff compliance to the minimum clinical requirement, means nonhealthcare-related demands, such as cybersecurity, get pushed down to a distant second position, Becker says.
But in its recently concluded hackathon, NTT Research said its Belgian team successfully demonstrated “a groundbreaking application” of ABE to protect images. ABE was introduced in 2005 in a paper by Brent Waters, NTT’s director of Cryptography and Information Security (CIS) Lab, and Amit Sahai, a professor of computer science at UCLA. It is a type of public-key encryption that allows for sharing data based on policies and attributes of the users — who the user is, rather than what they have.
Protecting DICOM Images With ABE
Essentially, ABE determines who can access data based on specific traits. ABE combines role-based encryption with content-based access and multi-authority access. For content-based access, ABE doesn’t just determine who gets access to data, but also what specific data they are allowed to access. Thus a radiologist might be able to access a CT scan but not patient identity, whereas a records clerk would be able to access identity but not imaging. Multiauthority access could come into play when a patient sees a specialist — the primary care physician might issue the specialist credentials to view a patient’s medical history, while a licensing board establishes credentials that allow them to write notes in that history; the specialist would need both sets of credentials to access the complete patient record.
The winning team’s three-part demo involved detecting and labeling a graphical object, encrypting the images and mapping between labels and ABE policies, and storing the objects, the metadata, and the blurred images in a database. Cabay’s coauthor, NTT senior software engineer Pascal Mathis, said their project uses an extract, transfer load (ETL) pipeline to transfer the images.
Mathis further explained that the AI component and encryption engine resides on an edge device, which sends only encrypted data to the database. Cabay says their project demonstrates how ABE can help to encrypt images in healthcare, such that “access is so locked-down that even the database administrator only sees images with blurred spots and encrypted information.”
Other major providers of picture archiving and communications systems (PACS), such as Philips, GE, and Sectra, are advancing solutions for digitization and increased automation of the imaging workflow, as part of a general migration to cloud-based systems and an enhanced security posture. These systems feature native end-to-end encryption and robust backup and breach prevention capabilities inherent to cloud environments. However, the DICOM data itself is not examined and may well be harboring malicious content, Becker notes.
“Standard detection-based network security tools, such as EDRs, XDRs, and MDRs, currently lack the capability to scan and disinfect DICOM imaging data,” he says. “It was this gap in security that moved us to develop, together with our healthcare partners, an imaging gateway that purifies the actual DICOM data stream itself.”
As healthcare becomes increasingly reliant on technology for more efficiency, healthcare industry leaders must prioritize using tools that enable the secure remote transmission of imaging studies to the hospitals’ PACS without incurring risk to the healthcare network.