New Paper Accepted: Rethinking Certification for Trustworthy Machine-Learning-Based Applications
Our paper entitled Rethinking Certification for Trustworthy Machine-Learning-Based Applications has been accepted for publication in the journal IEEE Internet Computing.
In this paper, we analyze the deficiencies of traditional certification schemes towards the certification of machine learning (ML)-based applications, and define a preliminary, novel certification scheme tailored for ML. It builds on the concept of multi-dimensional certification detailed in the paper Multi-Dimensional Certification of Modern Distributed Systems. Briefly, our we propose to evaluate the target of certification from different dimensions, namely the training set, the training process, and the resulting model. This approach permits to obtain a (preliminary) complete picture of the ML-based application to be certified.
The abstract is below.
Machine learning (ML) is increasingly used to implement advanced applications with nondeterministic behavior, which operate on the cloud-edge continuum. The pervasive adoption of ML is urgently calling for assurance solutions to assess applications’ nonfunctional properties (e.g., fairness, robustness, and privacy) with the aim of improving their trustworthiness. Certification has been clearly identified by policy makers, regulators, and industrial stakeholders as the preferred assurance technique to address this pressing need. Unfortunately, existing certification schemes are not immediately applicable to nondeterministic applications built on ML models. This article analyzes the challenges and deficiencies of current certification schemes, discusses open research issues, and proposes a first certification scheme for ML-based applications.
The paper is open access, go read it!