Rethinking Certification for Trustworthy Machine Learning-Based Applications
This addresses the need for trustworthy ML applications for policymakers, regulators, and industry, but is incremental as it builds on existing certification concepts.
The paper tackles the problem of certifying non-functional properties like fairness and robustness in machine learning applications, proposing a new certification scheme to address the inadequacy of existing methods for non-deterministic ML systems.
Machine Learning (ML) is increasingly used to implement advanced applications with non-deterministic behavior, which operate on the cloud-edge continuum. The pervasive adoption of ML is urgently calling for assurance solutions assessing applications non-functional properties (e.g., fairness, robustness, privacy) with the aim to improve their trustworthiness. Certification has been clearly identified by policymakers, regulators, and industrial stakeholders as the preferred assurance technique to address this pressing need. Unfortunately, existing certification schemes are not immediately applicable to non-deterministic 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.