The application of emerging technologies, such as Artificial Intelligence (AI), entails risks that need to be addressed to ensure secure and trustworthy socio-technical infrastructures. Machine Learning (ML), the most developed subfield of AI, allows for improved decision-making processes. However, ML models exhibit specific vulnerabilities that conventional IT systems are not subject to. As systems incorporating ML components become increasingly pervasive, the need to provide security practitioners with threat modeling tailored to the specific AI-ML pipeline is of paramount importance. Currently, there exist no well-established approach accounting for the entire ML life-cycle in the identification and analysis of threats targeting ML techniques. In this paper, we propose an asset-centered methodology—STRIDE-AI—for assessing the security of AI-ML-based systems. We discuss how to apply the FMEA process to identify how assets generated and used at different stages of the ML life-cycle may fail. By adapting Microsoft’s STRIDE approach to the AI-ML domain, we map potential ML failure modes to threats and security properties these threats may endanger. The proposed methodology can assist ML practitioners in choosing the most effective security controls to protect ML assets. We illustrate STRIDE-AI with the help of a real-world use case selected from the TOREADOR H2020 project.

Large-scale adoption of Artificial Intelligence and Machine Learning (AI-ML) models fed by heterogeneous, possibly untrustworthy data sources has spurred interest in estimating degradation of such models due to spurious, adversarial, or low-quality data assets. We propose a quantitative estimate of the severity of classifiers’ training set degradation: an index expressing the deformation of the convex hulls of the classes computed on a held-out dataset generated via an unsupervised technique. We show that our index is computationally light, can be calculated incrementally and complements well existing ML data assets’ quality measures. As an experimentation, we present the computation of our index on a benchmark convolutional image classifier.