Sébastien Mosser
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Sébastien Mosser has a total of 51 co-authors. Here are the top co-authors by number of shared publications:
- Richard F. Paige
McMaster University
5 shared publications - Alan Wassyng
McMaster University
3 shared publications - Benjamin Benni
Université Côte d’Azur, CNRS, I3S
2 shared publications - Mireille Blay-Fornarino
I3S Lab (UMR CNRS 7271), Université Côte d'Azur, Sophia Antipolis, France
2 shared publications - Nicolas Lacroix
I3S Lab (UMR CNRS 7271), Université Côte d'Azur, Sophia Antipolis, France
2 shared publications - Thomas Chiang
Computing and Software, McSCert, McMaster University, Hamilton, ON, Canada
2 shared publications
Latest Publications
- Open Access
- 4 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 813-822
A Model-Based Approach to Security Assessment of Robotic Systems in Healthcare: The Case of a Quadruped Robot
Robotic systems are increasingly being adopted in healthcare environments such as rehabilitation centres, in order to help support overburdened staff by performing more routine tasks. However, deploying a robot in a clinical setting introduces security risks that must be assessed before use. Small facilities rarely have access to the specialised expertise or budget needed to perform a full cybersecurity audit, and existing approaches to security analysis rely heavily on the expert knowledge of the analyst. We propose a model-based security analysis approach that assists non-experts in assessing the security of a deployed system. The approach guides non-experts in creating attack trees using both generic guidewords, as well as guidewords that are derived from system models. Additionally, an LLM is used to act as a security expert team member in the attack tree creation process. We have a security expert create a prompt template as a one-time activity, so that non-experts only need to fill in specific information on the system in order to get an attack tree output by the LLM. We use an example scenario of a Unitree Go2 Air robot dog deployed in a rehabilitation centre to demonstrate various aspects of our approach. We show initial steps towards creating a prompt template for the LLM, and show the usefulness of both the generic and model-based guidewords in creating more relevant attack trees by non-experts.
- 4 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 813-822
Open AccessA Model-Based Approach to Security Assessment of Robotic Systems in Healthcare: The Case of a Quadruped Robot
Robotic systems are increasingly being adopted in healthcare environments such as rehabilitation centres, in order to help support overburdened staff by performing more routine tasks. However, deploying a robot in a clinical setting introduces security risks that must be assessed before use. Small facilities rarely have access to the specialised expertise or budget needed to perform a full cybersecurity audit, and existing approaches to security analysis rely heavily on the expert knowledge of the analyst. We propose a model-based security analysis approach that assists non-experts in assessing the security of a deployed system. The approach guides non-experts in creating attack trees using both generic guidewords, as well as guidewords that are derived from system models. Additionally, an LLM is used to act as a security expert team member in the attack tree creation process. We have a security expert create a prompt template as a one-time activity, so that non-experts only need to fill in specific information on the system in order to get an attack tree output by the LLM. We use an example scenario of a Unitree Go2 Air robot dog deployed in a rehabilitation centre to demonstrate various aspects of our approach. We show initial steps towards creating a prompt template for the LLM, and show the usefulness of both the generic and model-based guidewords in creating more relevant attack trees by non-experts.
- Open Access
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 26-30
Model Cards for Responsible AI: Stop Carding, Start Modelling
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 26-30
Open AccessModel Cards for Responsible AI: Stop Carding, Start Modelling
- Open Access
Using Small Language Models to Reverse-Engineer Machine Learning Pipelines Structures
Background: Extracting the stages that structure Machine Learning (ML) pipelines from source code is key for gaining a deeper understanding of data science practices. However, the diversity caused by the constant evolution of the ML ecosystem (e.g., algorithms, libraries, datasets) makes this task challenging. Existing approaches either depend on non-scalable, manual labeling, or on ML classifiers that do not properly support the diversity of the domain. These limitations highlight the need for more flexible and reliable solutions. Objective: We evaluate whether Small Language Models (SLMs) can leverage their code understanding and classification abilities to address these limitations, and subsequently how they can advance our understanding of data science practices. Method: We conduct a confirmatory study based on two reference works selected for their relevance regarding current state-of-the-art's limitations. First, we compare several SLMs using Cochran's Q test. The best-performing model is then evaluated against the reference studies using two distinct McNemar's tests. We further analyze how variations in taxonomy definitions affect performance through an additional Cochran's Q test. Finally, a goodness-of-fit analysis is conducted using Pearson's chi-squared tests to compare our insights on data science practices with those from prior studies.
Open AccessUsing Small Language Models to Reverse-Engineer Machine Learning Pipelines Structures
Background: Extracting the stages that structure Machine Learning (ML) pipelines from source code is key for gaining a deeper understanding of data science practices. However, the diversity caused by the constant evolution of the ML ecosystem (e.g., algorithms, libraries, datasets) makes this task challenging. Existing approaches either depend on non-scalable, manual labeling, or on ML classifiers that do not properly support the diversity of the domain. These limitations highlight the need for more flexible and reliable solutions. Objective: We evaluate whether Small Language Models (SLMs) can leverage their code understanding and classification abilities to address these limitations, and subsequently how they can advance our understanding of data science practices. Method: We conduct a confirmatory study based on two reference works selected for their relevance regarding current state-of-the-art's limitations. First, we compare several SLMs using Cochran's Q test. The best-performing model is then evaluated against the reference studies using two distinct McNemar's tests. We further analyze how variations in taxonomy definitions affect performance through an additional Cochran's Q test. Finally, a goodness-of-fit analysis is conducted using Pearson's chi-squared tests to compare our insights on data science practices with those from prior studies.
- Open Access
- 22 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 725-734
A Tool For Feature-Requirement Traceability Using Requirement Canvas and Encapsulation
System traceability from requirements to implementation is desirable, often mandatory, for complex and safety-critical system development. However, maintaining traceability documentation can be extremely taxing and tedious through iterative and incremental development life cycles. The problem is further exacerbated when product variability is introduced in the system development. Feature modelling and product line engineering are becoming increasingly common in safety-critical domains such as the automotive domain when manufacturers must consider maintaining documentation for multiple variations of a single vehicle. This problem is made even more complex by the yearly iterations they have on their vehicles. Thus, we introduce CyclicL, a tool for developing, managing, and modelling feature-requirement traceability through iterative and incremental development strategies. CyclicL is designed to facilitate traceability between features and requirements through various product iterations, and multiple incremental feature and requirement changes.
- 22 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 725-734
Open AccessA Tool For Feature-Requirement Traceability Using Requirement Canvas and Encapsulation
System traceability from requirements to implementation is desirable, often mandatory, for complex and safety-critical system development. However, maintaining traceability documentation can be extremely taxing and tedious through iterative and incremental development life cycles. The problem is further exacerbated when product variability is introduced in the system development. Feature modelling and product line engineering are becoming increasingly common in safety-critical domains such as the automotive domain when manufacturers must consider maintaining documentation for multiple variations of a single vehicle. This problem is made even more complex by the yearly iterations they have on their vehicles. Thus, we introduce CyclicL, a tool for developing, managing, and modelling feature-requirement traceability through iterative and incremental development strategies. CyclicL is designed to facilitate traceability between features and requirements through various product iterations, and multiple incremental feature and requirement changes.
- Open Access
- 22 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 1025-1034
Building deduplicated model repositories to assess domain-specific languages evolution
Software evolution and maintenance is a real challenge in modern software engineering. In the context of model-driven development, which heavily rely on interconnected (meta-)models, tools and generators, evolving both models and their associated meta-models is particularly complex. This issue is also prevalent in language engineering, where evolving a language's grammar or semantics must remain consistent with the pre-existing models. In this paper, we explore how techniques inspired by repository mining can help a model designer/language engineer to build a deduplicated dataset of existing models available in open source repositories. Deduplication is essential to ensure the evolution made on the meta-model/language can be efficiently assessed. We apply the method to the P4 language, an industrial domain-specific language (Intel, Linux foundation) used to model software defined networks.
- 22 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 1025-1034
Open AccessBuilding deduplicated model repositories to assess domain-specific languages evolution
Software evolution and maintenance is a real challenge in modern software engineering. In the context of model-driven development, which heavily rely on interconnected (meta-)models, tools and generators, evolving both models and their associated meta-models is particularly complex. This issue is also prevalent in language engineering, where evolving a language's grammar or semantics must remain consistent with the pre-existing models. In this paper, we explore how techniques inspired by repository mining can help a model designer/language engineer to build a deduplicated dataset of existing models available in open source repositories. Deduplication is essential to ensure the evolution made on the meta-model/language can be efficiently assessed. We apply the method to the P4 language, an industrial domain-specific language (Intel, Linux foundation) used to model software defined networks.
- Open Access
- 22 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 1180-1187
Mapping Requirements to Features to Create Traceability in Product Line Models
Developers and engineers have shown the many benefits of feature modelling for identifying reusable software and hardware components across product lines and platforms. However, a gap exists when it comes to identifying reusable products, and requirements across iterations of features. Due to the ambiguous relationship between features and requirements, it can be difficult to determine the boundary between them. We propose a methodology for using features to encapsulate requirements by establishing a clear hierarchy between features and requirements. This will assist feature-requirement traceability tooling and automation development.
- 22 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 1180-1187
Open AccessMapping Requirements to Features to Create Traceability in Product Line Models
Developers and engineers have shown the many benefits of feature modelling for identifying reusable software and hardware components across product lines and platforms. However, a gap exists when it comes to identifying reusable products, and requirements across iterations of features. Due to the ambiguous relationship between features and requirements, it can be difficult to determine the boundary between them. We propose a methodology for using features to encapsulate requirements by establishing a clear hierarchy between features and requirements. This will assist feature-requirement traceability tooling and automation development.
- Open Access
- 3 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Software
- Vol. 41 (4) , 124-133
Polyglot Software Development: Wait, What?
We propose a concise and unambiguous definition of polyglot software development with a conceptual model and characterize the techniques used for the specification and operationalization of polyglot software development with a feature model.
- 3 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Software
- Vol. 41 (4) , 124-133
Open AccessPolyglot Software Development: Wait, What?
We propose a concise and unambiguous definition of polyglot software development with a conceptual model and characterize the techniques used for the specification and operationalization of polyglot software development with a feature model.
- 1 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 122-129
Towards Personalized Learning Paths to Empower Competence Development in Model Driven Engineering Through the ENCORE Platform
Developing engaging learning experiences is costly and complicated. Open Educational Resources (OERs) offer the possibility of reuse, creating opportunities for more efficient and effective development of learning, including the potential for truly personalised learning through adaptive learning management platforms. To make this a reality, we need to address two challenges: (1) for OERs to become effective there need to be tools that allow them to be reused efficiently from high-level designs of learning experiences, and (2) for such high-level tools to work, we need to establish a robust infrastructure that treats OERs as components, akin to software components, complete with well-defined interfaces. The latter is particularly challenging for learning resources on MDE, because of the often complex tool environments required. In this paper, we propose a platform for personalized learning design based on OERs and provide initial insights into component formats within the context of teaching about model-driven engineering.
- 1 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 122-129
Towards Personalized Learning Paths to Empower Competence Development in Model Driven Engineering Through the ENCORE Platform
Developing engaging learning experiences is costly and complicated. Open Educational Resources (OERs) offer the possibility of reuse, creating opportunities for more efficient and effective development of learning, including the potential for truly personalised learning through adaptive learning management platforms. To make this a reality, we need to address two challenges: (1) for OERs to become effective there need to be tools that allow them to be reused efficiently from high-level designs of learning experiences, and (2) for such high-level tools to work, we need to establish a robust infrastructure that treats OERs as components, akin to software components, complete with well-defined interfaces. The latter is particularly challenging for learning resources on MDE, because of the often complex tool environments required. In this paper, we propose a platform for personalized learning design based on OERs and provide initial insights into component formats within the context of teaching about model-driven engineering.
- 14 September 2023
- Published by Springer Nature
- p. 5-14
Using Assurance Cases to Prevent Malicious Behaviour from Targeting Safety Vulnerabilities
- 14 September 2023
- Published by Springer Nature
- p. 5-14
Using Assurance Cases to Prevent Malicious Behaviour from Targeting Safety Vulnerabilities
- 15 September 2022
- Published by Springer Nature
- p. 247-261
Automatically Estimating the Severity of Multiple Symptoms Associated with Depression
- 15 September 2022
- Published by Springer Nature
- p. 247-261
Automatically Estimating the Severity of Multiple Symptoms Associated with Depression