Towards Certification of Machine Learning-Based Distributed Systems - Challenges and a First Scheme for Non-Functional Properties

Aug 4, 2026

Upload by Himbo

Machine Learning is increasingly used to operate complex distributed systems across the cloud–edge continuum enabled by 5G, making system behavior more non-deterministic. Certification, a leading technique for verification of non-functional properties, cannot be directly applied to systems driven by ML inference. The work examines gaps and challenges in existing certification schemes and highlights the regulatory and industrial demand for assurance of properties such as fairness, robustness, and privacy. It then proposes an initial certification approach for ML-based distributed systems.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages5

Reading Time13 min

AccessFree to download

Introduction

Why certification is changing for ML-driven distributed systems

Motivation for certifying ML behavior and non-functional properties

Main certification scheme proposal

Towards Certification of Machine Learning-Based Distributed Systems Marco Anisetti, Claudio A. Ardagna, Nicola Bena, Ernesto Damiani arXiv :2305 . 16822v2 [ cs .LG] 1 Jun 2023 Abstract—Machine Learning (ML) is increasingly used to drive the operation of complex distributed systems deployed on the cloud-edge continuum enabled by 5G. Correspondingly, distributed systems’ behavior is becoming more non-deterministic in nature. This evolution of distributed systems requires the definition of new assurance approaches for the verification of non-functional properties. Certification, the most popular assurance technique for system and software verification, is not immediately applicable to systems whose behavior is determined by Machine Learning-based inference. However, there is an increasing push from policymakers, regulators, and industrial stakeholders towards the definition of techniques for the certification of non-functional properties (e.g., fairness, robustness, privacy) of ML. This article analyzes the challenges and deficiencies of current certification schemes, discusses open research issues and proposes a first certification scheme for ML-based distributed systems. ~~ ~~ ✦ ~~ ~~ 1 INTRODUCTION Modern distributed systems are composed of elastic serverside processes running in the cloud, consisting of microand nano-services developed with cloud-native technologies and orchestrated at run time. The availability of new orchestration platforms and programming frameworks designed for them is making it possible to perform network computing at line speed [1] . In the fullness of time, this cloud continuum will enable autonomous, highly intelligent systems,including innovative hardware devices [2] . While these systems promise undeniable advantages in terms of availability, elasticity, and, ultimately, quality of service, their complexity is much higher than their predecessors’. This complexity leap is affecting the governance, risk, and compliance landscape and even the procedures to guarantee safety and security of citizens. Complex distributed systems call for novel solutions to assess and verify their behavior. To understand why, let us discuss were we stand right now. Security assurance is the set of techniques to evaluate and increase the trustworthiness of a distributed system [3] . Along the years, certification has become the most popular assurance technique, providing a way for a trusted authority to assert that a system supports a given (set of) non-functional properties according to some trust evidence of its operation. Certificates enable users to make informed decisions about using a system on the basis of the behavior guaranteed by the accompanying certificates. Certification schemes have been applied to software systems since the 80s with the definition of the Orange book, and then applied to software systems, services, and cloud. These test-based schemes underwent a crisis in the middle of the past decade, when it became clear that certification of distributed, service-based applications required run-time reverification as potentially different services were recruited at • Marco Anisetti, Claudio A. Ardagna, Nicola Bena, Ernesto Damiani are with the Department of Computer Science, Università degli Studi di Milano, Milan, Italy. • Ernesto Damiani is alo with with Khalifa University of Science and Technology, Abu Dhabi, UAE. each execution. In 2012, on the crest of the service-oriented computing wave, the seminal ASSERT4SOA project1 proposed a new generation of dynamic certification techniques with the following slogan: You live in a certified house, you drive a certified car, why would you use an uncertified service. Today, a second crisis of certification schemes is looming, as the advent of machine learning is radically transforming the notion of orchestration. Distributed systems are increasingly becoming non-deterministic as their behavior is driven by ML models, which replace deterministi

FAQs

Why traditional certification schemes are difficult to apply to ML-based distributed systems?

ML-driven orchestration makes system behavior non-deterministic and dependent on inference results, often from black-box models. As a result, classic certification assumptions and verification evidence methods do not directly transfer.

Which non-functional properties are highlighted as targets for ML certification?

The document emphasizes fairness, robustness, privacy, and related safety and reliability properties. These properties are framed as outcomes guaranteed by certification evidence, not as purely theoretical properties of the models.

What factors make certified guarantees for ML-based systems challenging?

Training data issues such as partial or inaccurate data, poisoned data, and sensitive data can respectively harm fairness, robustness/security, and privacy/explainability. The surrounding deployment context also affects the resulting properties.

Aug 4, 2026

Upload by Himbo