In the dynamic landscape of network security, where cyberattacks continuously evolve, robust and adaptive detection mechanisms are essential, particularly for safeguarding Internet of Things (IoT) networks. This paper introduces an advanced anomaly detection model that utilizes Artificial Intelligence (AI) to identify network anomalies based on traffic features, explaining the most influential factors behind each detected anomaly. The model integrates domain knowledge stored in a knowledge graph to verify whether the detected anomaly constitutes a legitimate attack. Upon validation, the model identifies which core cybersecurity principles—Confidentiality, Integrity, or Availability (CIA)—are violated by mapping influential feature values. This is followed by an alignment with the MITRE ATT&CK framework to provide insights into potential attack tactics, techniques, and intelligence-driven countermeasures. By leveraging explainable AI (XAI) and incorporating expert domain knowledge, our approach bridges the gap between complex AI predictions and human-understandable decision-making, thereby enhancing both detection accuracy and result interpretability. This transparency facilitates faster responses and real-time decision-making while improving adaptability to new, unseen cyber threats. Our evaluation on network traffic datasets demonstrates that the model not only excels in detecting and explaining anomalies but also achieves an overall detection accuracy of 0.97 with the integration of domain knowledge for attack legitimacy. Furthermore, it provides 100% accuracy for threat intelligence based on the MITRE ATT&CK framework, ensuring that security measures are verifiable, actionable, and ultimately strengthen IoT environment defenses by delivering real-time threat intelligence and responses, thus minimizing human response time.

Machine learning and Artificial Intelligence (AI) already support human decision-making and complement professional roles, and are expected in the future to be sufficiently trusted to make autonomous decisions. To trust AI systems with such tasks, a high degree of confidence in their behaviour is needed. However, such systems can make drastically different decisions if the input data is modified, in a way that would be imperceptible to humans. The field of Adversarial Machine Learning studies how this feature could be exploited by an attacker and the countermeasures to defend against them. This work examines the Fast Gradient Signed Method (FGSM) attack, a novel Single Value attack and the Label Flip attack on a trending architecture, namely a 1-Dimensional Convolutional Neural Network model used for time series classification. The results show that the architecture was susceptible to these attacks and that, in their face, the classifier accuracy was significantly impacted.

Deep learning approaches have been established as the main methodology for video classification and recognition. Recently, 3-dimensional convolutions have been used to achieve state-of-the-art performance in many challenging video datasets. Because of the high level of complexity of these methods, as the convolution operations are also extended to an additional dimension in order to extract features from it as well, providing a visualization for the signals that the network interpret as informative, is a challenging task. An effective notion of understanding the network's innerworkings would be to isolate the spatio-temporal regions on the video that the network finds most informative. We propose a method called Saliency Tubes which demonstrate the foremost points and regions in both frame level and over time that are found to be the main focus points of the network. We demonstrate our findings on widely used datasets for thirdperson and egocentric action classification and enhance the set of methods and visualizations that improve 3D Convolutional Neural Networks (CNNs) intelligibility. Our code1and a demo video2are also available.

The main challenge of exposing IEC61499 or IEC61131-3 Function Blocks as a service remains in adopting service-oriented concepts in function block programming. Implementing an IEC61499 or IEC61131-3 Function Block that is being accessed via service-oriented protocols is straight forward. The main challenge remains in modeling a Function Block as a service. Adopting service-oriented concepts like Service Oriented Architecture or Microservice Architecture requires tackling challenges like service granularity, (de)composition, etc. For instance, too coarse-grained services could lead to significant drawbacks, while too fine-grained services could increase the system's overall complexity, introducing semantic tight coupling and bringing about communication overhead. Therefore, understanding whether a service (de)composition is adding any value could help us to identify the best service granularity. Finding the best service granularity means knowing how many Function Blocks could be exposed into one service. This could undeniable lead to improvement in resource consumption especially in constraint environments. In this paper we design a cost analysis function for calculating the overhead of service decomposition. This work will help to answer one of the most important aspects of the service-oriented approach, called service granularity in the scope of Function Block as a Service.

Research in adaptive learning is mainly focused on improving learners’ learning achievements based mainly on personalization information, such as learning style, cognitive style or learning achievement. In this paper, an innovative adaptive learning approach is proposed based upon two main sources of personalization information that is, learning behaviour and personal learning style. To determine the initial learning styles of the learner, an initial assigned test is employed in our approach. In order to more precisely reflect the learning behaviours of each learner, the interactions and learning results of each learner are thoroughly recorded and in depth analysed, based on advanced machine learning techniques, when adjusting the subject materials. Based on this rather innovative approach, an adaptive learning prototype system has been developed.

TIEC 61499 defines an execution model for distributed industrial control applications, i.e. a single application distributed among several devices. In such an environment partial failures are likely to occur. In order to avoid probable system malfunctions and breakdowns due to partial failures, the authors have previously proposed a framework where the concept of replication may be applied to the IEC 61499 execution model. This paper focuses on describing an implementation of this replication framework on the FORTE IEC 61499 execution platform, along with the results of the first tests of the implementation. A set-up for the full validation of the approach is also described.

In a modern industrial environment control programs are distributed among several devices. This raises new issues and challenges especially in failure modes. Building fault tolerant applications can be the solution in order a failure of one sub-component not to jeopardize the execution of the whole application. The authors have proposed a framework to support replicated IEC 61499 applications. In this paper we augment this framework with the support for different voting strategies, propose an extension of the replication communication protocol, and analyse the resulting fault-tolerance semantics. A limited implementation of the framework is also described.

Industrial processes use energy to transform raw materials and intermediate goods into final products. Many efforts have been done on the minimization of energy costs in industrial plants. Apart from working on “how” an industrial process is implemented, it is possible to reduce the energy costs by focusing on “when” it is performed. Although, some manufacturing plants (e.g. refining or petrochemical plants) can be inflexible with respect to time due to interdependencies in processes that must be respected for performance and safety reasons, there are other industrial segments, such as alumina plants or discrete manufacturing, with more degrees of flexibility. These manufacturing plants can consider a more flexible scheduling of the most energy-intensive processes in response to dynamic prices and overall condition of the electricity market. In this scenario, requests for energy can be encoded by means of a formal structure called flex-offers, then aggregated (joining several flex-offers into a bigger one) and sent to the market, scheduled, disaggregated and transformed into consumption plans, and eventually, into production schedules for given industrial plant. In this paper, we describe the flex-offer concept and how it can be applied to industrial and home automation scenarios. The architecture proposed in this paper aims to be adaptable to multiples scenarios (industrial, home and building automation, etc.), thus providing the foundations for different concept implementations using multiple technologies or supporting various kinds of devices.

The Arrowhead project aims to address the technical and applicative issues associated with cooperative automation based on Service Oriented Architectures. The problems of developing such kind of systems are mainly due to the lack of adequate development and service documentation methodologies, which would ease the burden of reusing services on different applications. The Arrowhead project proposes a technical framework to efficiently support the development of such systems, which includes several tools for documentation of services and to support the development of SOA-based installations. The work presented in this paper describes the approach which has been developed for the first generation pilots to support the documentation of their structural services. Each service, system and system-of-systems within the Arrowhead Framework must be documented and described in such way that it can be implemented, tested and deployed in an interoperable way. This paper presents the first steps of realizing the Arrowhead vision for interoperable services, systems and systems-of-systems.