Academics are developing new methods to protect AI-enabled hospital diagnostic tools

Published: 3-Sep-2024

Academics at the University of Huddersfield are developing a new method to protect AI-enabled diagnostic tools used in hospitals from cyber-attacks

Dr Faisal Jamil, a lecturer in computing at the Centre for Autonomous and Intelligent Systems department at the University of Huddersfield, is spearheading the creation of new software which is being designed specifically for healthcare AI security and patient data privacy.

This comes at a time when Synnovis, a company that supplies lab services such as blood tests, swabs, and bowel tests to major London hospitals, was hit by a ransomware cyber attack in June. 

CyberASAP, a UK cyber security academic startup accelerator programme, provided funding for the Secure Threat Intelligence Sharing Platform software.

Dr Jamil's team has reached the second phase of the highly competitive national programme. The members of the team include:

  • Dr Saad Khan, Senior Lecturer in cyber security
  • Professor Simon Parkinson, Professor of cyber security and Director of the University’s Centre for Cybersecurity
  • Farah Barbar, a Researcher in the Department of Computer Science

Our platform is designed for hospitals that are adapting AI for diagnosis and treatment plans

After successfully completing both the market validation and proof of concept rounds, they have to date received funding of £32,000 alongside support to turn their concept into reality.

Next, the team will be presenting their work to Innovate UK at the end of July in a bid to secure further funds of £60,000 to develop it through to a product ready for the commercial market.

Dr Jamil said: "Our platform is designed for hospitals that are adapting AI for diagnosis and treatment plans. Securing these AI models is critical as they are vulnerable to attacks and traditional security can’t stop such attacks on AI models."

"We offer a secure platform – our platform uses federated learning and swarm intelligence to protect sensitive patient data. We would also share threat intelligence anonymously across the healthcare network to detect and stop attacks in real-time," Dr Jamil continued.

Securing these AI models is critical as they are vulnerable to attacks and traditional security can’t stop such attacks on AI models

Dr Jamil added that the Secure Threat Intelligence Sharing Platform would protect the integrity of the AI-enabled diagnostic tool, as such healthcare tech becomes more commonplace at a time when hospitals face an increasing number of cyber-attacks such as data poisoning and the use of evil models.

The Secure Threat Intelligence Sharing Platform also offers reliable and consistent performance, keeping patient data secure as well as a collaborative advantage with improved security through using collective insights to stop attacks before they happen.

The team have identified pharmaceutical companies, research institutions, as well as hospitals and clinics as potential first customers for the platform.

This includes early disease detection and imaging analysis, such as the AI-powered red dot chest X-ray system currently used by Calderdale and Huddersfield NHS Foundation Trust to assess chest scans in a quick turnaround reducing waiting times for patients.

The programme creates a pipeline to move great cyber security ideas out of the university lab and into the commercial market

Dr Jamil explained that while their current focus is AI-enabled diagnostic tools in healthcare, the platform could be used by any company utilising an artificial intelligence model.

The CyberASAP programme provides academics with the expertise, knowledge and training needed to convert their research into technologies, products and services.

Funded by the Government’s Department for Science, Innovation, and Technology and delivered by Innovate UK, the programme creates a pipeline to move great cyber security ideas out of the university lab and into the commercial market.

Trending Articles

You may also like