Using attack graphs to defend healthcare systems from cyberattacks: a longitudinal empirical study

Area of Science:

• Cybersecurity
• Healthcare IT
• Machine Learning

Background:

• Hospitals are highly vulnerable to cyberattacks, jeopardizing sensitive patient and financial data.
• Protecting medical records is paramount due to significant personal and financial implications.
• Attack graphs offer a potential method for securing healthcare systems.

Purpose of the Study:

• To statistically examine real-life cyberattacks on healthcare institutions.
• To model Industrial Control Systems (ICS) vulnerability data using machine learning.
• To identify trends and predict future cyber threats in healthcare IT systems.

Main Methods:

• Statistical analysis of 352 real-life cyberattacks using Common Vulnerability Scoring System (CVSS) data.
• Application of machine learning techniques and an artificial neural network to model ICS vulnerability data.
• Utilized Python programming for developing and enhancing the K-nearest neighbors (KNN) algorithm.

Main Results:

• Healthcare IT systems exhibit very high average vulnerability scores.
• Vulnerability scores are significantly higher in institutions with a history of cyberattacks and no implemented mitigation.
• The K-nearest neighbors (KNN) algorithm proved most effective in modeling healthcare cyberattacks.

Conclusions:

• Medical records are at critical risk due to a high likelihood of cyberattacks in healthcare institutions.
• Urgent precautionary measures are essential to mitigate cyber risks and enhance the security of healthcare systems.
• Institutions must prioritize robust cybersecurity to ensure reliability and robustness against threats.