This article deals with scheduling anomalies in real-time systems. We present MONANO, a POSIX user-level library allowing applications to dynamically detect a preidentified set of real-time scheduling anomalies. The MONANO library is based on the modelling of architecture and runtime constraints. MONANO monitors during the runtime the timing behavior of the application and deduces properties needed to identify scheduling anomalies. We present also a benchmark to evaluate our approach. The benchmark is composed of several programs implementing the most frequent real-time scheduling anomalies.

The development of real-time systems is one of the areas with the highest relevance in computer science, and the number of critical systems has increased significantly. These systems considers several applications running concurrently, and inside each of those applications code might be parallelized to improve their performance and control the priority of each parallelizable task. Several efforts have been done in different programming languages to provide real-time systems with parallel programming models, whether by code extensions or annotations, or with specific features in the actual language core.

Smart mobility is emerging, addressing heterogeneous scenarios with high impact on technology infrastructures, solutions, and people. Safety and availability are mandatory, forcing the design of new reliable services for localization, health monitoring of the user, maintenance of vehicle, and protection of the environment. This paper proposes a container-based microservice approach to the edge computing in IoT smart mobility scenarios. Since smart mobility backends must manage a large heterogeneity of applications, the proposed approach is promising with respect to the classical solutions (based on "monolithic hardware+software" devices), from the point of view of flexibility, upgradability, security, scalability, and reliability. A demo use case, based on industry-grade hardware and Docker, has been realized and multiple implementations of the same services have been executed in parallel, showing strong independence between them. Moreover, average delays of less than 10 ms are obtained, confirming the usability in several smart mobility (and smart city) applications.

Machine Learning (ML) models are increasingly used in systems that involve physical human interaction or decision-making systems that impact human health and safety. Ensuring that these systems are safe and reliable is an important topic of current AI research. For many ML models it is unclear how a prediction (output) is arrived at from the provided features (input). Critical systems cannot blindly trust the predictions of such "black box" models, but instead need additional reassurance via insight into the model's reasoning. A range of methods exist within the field of Explainable AI (XAI) to make the reasoning of black box ML models more understandable and transparent. The explanations provided by XAI methods may be evaluated in a number of (competing) ways. In this paper, we investigate the trade-off between selected metrics for an XAI method called UnRAvEL, which is similar to the popular LIME approach. Our results show that by weighting the terms within the acquisition function used in UnRAvEL, different trade-offs can be achieved.

The MORPHEMIC project covers several features from modelling cross-cloud applications, continuous and autonomous optimization and deployment and providing access to several cloud capabilities for data intensive application needing on one hand more and more resources, and, on the other hand, to support different models, such as Edge and Fog, besides the Cloud.

To help automatize the security risk assessment process of Cyber-Physical Systems (CPS), we propose a tool based on Discrete Event Systems (DES) to model the architecture and the behavior of CPS in the presence of cyberattacks. Then, we present a lightweight algorithm to generate all the attack scenarios threatening a system, i.e. the sequences of attacks leading to a critical state (e.g. loss of control, collision, etc.). This kind of generation being prone to combinatorial explosion, our algorithm embeds state-space reduction capabilities focused on the specificities of cyber-physical attacks. Finally, we illustrate the performance of our algorithm on a case study: the navigation system of an autonomous vessel. This work can be seen as an alternative to heavy tools expressed in specific languages. It is open source and aims to give a good compromise between expressiveness, modeling time and computational power.

Deep neural networks (DNNs) have demonstrated promising performances in handling complex real-world scenarios, surpassing human intelligence. Despite their exciting performances, DNNs are not robust against adversarial attacks. They are specifically vulnerable to data poisoning attacks where attackers meddle with the initial training data, despite the multiple defensive methods available, such as defensive distillation. However, defensive distillation has shown promising results in robustifying image classification deep learning (DL) models against adversarial attacks at the inference level, but they remain vulnerable to data poisoning attacks. This work incorporates a data denoising and reconstruction framework with a defensive distillation methodology to defend against such attacks. We leverage a denoising autoencoder (DAE) to develop a data reconstruction and filtering pipeline with a well-designed reconstruction threshold. We added carefully created adversarial examples to the initial training data to assess the proposed method's performance. Our experimental findings demonstrate that the proposed methodology significantly reduced the vulnerability of the defensive distillation framework to a data poison attack.

Nowadays, the increasing complexity of Advanced Driver Assistance Systems (ADAS) and Automated Driving (AD) means that the industry must move towards a scenariobased approach to validation rather than relying on established technology-based methods. This new focus also requires the validation process to take into account Safety of the Intended Functionality (SOTIF), as many scenarios may trigger hazardous vehicle behaviour. Thus, this work demonstrates how the integration of the SOTIF process within an existing validation tool suite can be achieved. The necessary adaptations are explained with accompanying examples to aid comprehension of the approach.

Due to the increasing performance demands of missionand safety-critical Cyber-Physical Systems (of Systems), these systems exhibit a rapidly growing complexity, manifested by an increasing number of (distributed) computational cores and application components connected via complex networks. However, with these systems' growing complexity and interconnectivity, the chances of hardware failures and disruptions due to cyber-attacks will also quickly increase.

Cyber-physical systems are usually integrated with a highly changing environment. Because of the variability of their environment, these systems need to adapt their behaviour at runtime to achieve the expected results. In the case of distributed solutions, this behaviour must be consistent across the continuum, increasing the complexity of the design due to the need to synchronise the different components distributed in the multiple nodes. A multimode application approach could be used to solve this need, assuming it is implemented as a distributed system. The approach involves applications with multiple behaviours (or operational modes) and a mechanism to change from one mode to another. This article describes an operational mode change management solution based on that approach and the reference architecture presented in the TRANSACT project. This architecture integrates CPS end devices at the edge of the network with edge computing servers and cloud computing facilities. Static and dynamic views are included to describe the solution, and specific use case scenarios are added for better understanding.