SafetyMed: Protecting Connected Medical Devices from Hackers with AI

By Nuruzzaman Faruqui

·

Heart monitors, insulin pumps and hospital sensors are increasingly connected to the internet, and that makes them targets. SafetyMed is an AI security guard designed to catch hackers trying to break into these medical devices.

97.63%average accuracy detecting intrusions
2kinds of attack data checked: network traffic and images
FirstIDS to protect medical IoT from both

The problem

The Internet of Medical Things (IoMT) is growing fast and holds valuable health data, which attracts cybercriminals. But these devices are built to be cheap and to use little power, so they cannot run the heavy security software a laptop can.

The result is a wave of intrusions into devices that patients’ health may depend on.

What the research did

We designed SafetyMed, an intrusion detection system that sits on the network and watches for attacks. It combines two types of AI: a convolutional neural network, which is good at spotting patterns in images and grids of data, and a long short-term memory (LSTM) network, which is good at understanding sequences over time, like the flow of network traffic.

That combination lets SafetyMed catch attacks hidden in malicious image data as well as in suspicious network traffic. We also tuned it carefully to balance catching real attacks against raising false alarms.

What we found

SafetyMed detected intrusions with an average accuracy of 97.63%, precision of 98.47%, recall of 97% and an F1-score of 97.73%. It is the first intrusion detection system designed to protect medical IoT devices from both malicious images and malicious traffic sequences.

Who benefits

  • Patients who rely on connected devices for monitoring and treatment
  • Hospitals, which hold sensitive data and run networks full of low-power devices
  • Device makers, who can add protection at the network level without redesigning cheap hardware

Limits and what comes next

SafetyMed was evaluated on research datasets. Attackers constantly change tactics, so any real system needs ongoing updates and testing on live hospital networks. My later work, including IntruSafe, extends this line of research.

The team

I led this research with Mohammad Abu Yousuf, Md Whaiduzzaman (QUT), AKM Azad and Salem A. Alyami (Imam Mohammad Ibn Saud Islamic University), Pietro Liò (University of Cambridge), Muhammad Ashad Kabir and Mohammad Ali Moni (Charles Sturt University).

Frequently asked questions

Why are medical devices easy to hack?

They are designed to be low-cost and low-power, which leaves little room for strong built-in security. That makes protection at the network level important.

What is an intrusion detection system?

It is software that watches network activity and raises an alarm, or blocks traffic, when it spots signs of an attack.

How accurate is SafetyMed?

It detected intrusions with 97.63% average accuracy in testing.

About the author

Nuruzzaman Faruqui is an award-winning AI educator and researcher who has taught artificial intelligence and machine learning at university level since 2018. A two-time winner of the Excellence in Teaching Award (2023, 2024) and winner of the Best Researcher Award (2024) at Daffodil International University, he is now a doctoral researcher in AI at Monash University, Australia. He has published 53 papers and six books, and writes these articles to make AI research easy for everyone to understand. His teaching · About Zaman · Get in touch

Keep reading