
Technologies
From research to engineering practice
What a DEIS technology is
A DEIS technology is a method that has left the paper it was published in. Each one is software an engineer can run on their own system, with documentation, examples and a route to a licence or support. Each grew out of the group’s research themes, and each is still developed as that research continues.
Group leads: Prof. Yiannis Papadopoulos, Lead · Dr Koorosh Aslansefat, Co-Lead
Contact · Prof. Yiannis Papadopoulos
Hierarchically Performed Hazard Origin and Propagation Studies: generates fault trees and FMEAs from an annotated system model, allocates safety requirements, and optimises designs for dependability and cost.
Status: in industrial use, with commercial and research licences
Lead · Dr Koorosh Aslansefat
Runtime safety monitoring for machine learning classifiers: measures how far live input data has drifted from the training data, so a classifier can flag when it is outside the conditions it was validated for.
Status: open source under the MIT licence, in Python, MATLAB and R
Projects · SESAME, and autonomous shipping with NYK
Executable Digital Dependability Identities: selected safety knowledge made executable, so a system can monitor itself, reassess its safety guarantees while it runs, and share them with the systems it cooperates with.
Status: research toolchain for robotics and autonomous Systems of Systems
Lead · Dr Koorosh Aslansefat
Explainable AI with SMILE: explains individual decisions of black-box models, from classifiers and image models to large language models, using statistical distance measures to make local explanations more reliable.
Status: open source under the MIT licence, installed with pip install xwhy
How the technologies connect
The four technologies cover a system from design to operation, and three of them came together in the SESAME project.
Design: HiP-HOPS generates fault trees and FMEAs from the system architecture.
Learned components: SafeML checks whether a classifier’s inputs still match its training data, and XWhy explains why a model made a particular decision.
Operation: EDDIs execute the safety model and share dependability guarantees at runtime, drawing on HiP-HOPS models and SafeML monitors.

Want to use one of these tools?
We support evaluations, research licences and funded tool development for each of the four technologies.