As the adoption of electric vehicles (EVs) accelerates, ensuring the security of the EV charging infrastructure has become a critical challenge in the field of cyber-physical system (CPS) security. The EV charging system represents a complex CPS environment where physical and digital components are intricately coupled with intertwined layered communication protocols. These characteristics necessitate a comprehensive analysis of threat flows rather than merely identifying fragmented vulnerabilities. However, conventional threat modeling techniques often rely on static system analyses, providing threat checklists or security requirement specifications, limiting their ability to capture realistic attack paths and stepwise adversarial behaviors. More dynamic threat modeling approaches that incorporate attacker behavior flows are needed to address this limitation. This paper proposes a threat modeling method that analyzes system operation flows and asset interactions to identify critical attack points. This work constructs an attack tree based on this analysis and applies behavior-based threat classification at the leaf-node level. The resulting threat scenarios are linked to the MITRE adversarial tactics, techniques, and common knowledge framework with flows to generate actionable and executable attack scenarios. The proposed method is applied to a core component of the EV charging infrastructure (i.e., EV supply equipment) to derive scenarios, including remote intrusions, physical access attacks, and supply chain compromises. The scenarios are designed to be directly executable in breach and attack simulation environments, enabling automated and repeatable validation of security controls and facilitating proactive defense planning.
Connected and automated vehicles rely on data from various entities to support safety-critical applications such as Cooperative Adaptive Cruise Control (CACC). However, unauthorized data manipulation through, for example, data injection attacks can compromise vehicle safety and lead to incidents. Existing vehicular security mechanisms, such as Misbehavior Detection System (MBD), have limitations in detecting and mitigating all types of threats on their own. To address these limitations, our prior work has proposed the concept of a Trust Assessment Framework (TAF), which assesses data trustworthiness by combining evidence from multiple security systems operating as trust sources. However, TAF as a concept has not been extensively evaluated in safety-critical Cooperative Driving (CD) applications. In this work, we refine the architecture of the TAF and implement a software prototype based on it. We integrate the TAF prototype with a CACC simulation environment and implement three types of data injection attacks. We demonstrate that by incorporating multiple security mechanisms as trust sources, the TAF significantly improves attack detection performance and reduces the number of crashes by 86% compared to using a single security mechanism, such as MBD.
Modern vehicles increasingly depend on wireless technologies such as Remote Keyless Entry (RKE) systems for user convenience. However, these systems are susceptible to various security vulnerabilities. The RollJam attack, for example, demonstrated that rolling code mechanisms, intended to prevent unauthorized access, can be circumvented through wireless interception and replay techniques. In addition, attackers have exploited access to the Controller Area Network (CAN) bus to carry out a range of in-vehicle attacks. In this paper, we introduce RollCAN, a novel attack that adapts the principles of RollJam from the wireless domain to the CAN bus. RollCAN enables adversaries with physical access to intercept, block, and replay RKE messages over the CAN bus, thereby allowing unauthorized locking and unlocking of vehicle doors. We demonstrate the practicality of the RollCAN attack through two distinct implementations: one using professional automotive-grade equipment and another using a low-cost Raspberry Pi-based setup with a CAN interface. Furthermore, we discuss potential mitigation strategies to counteract the RollCAN attack.
Driven by technological progress and, above all, software, our world is becoming ever faster, more flexible and more complex. Vehicles in particular are complex systems due to the number of components, communication, and millions of lines of code. Recent trends such as autonomous driving and artificial intelligence further accelerate these developments. As a result, the attack surface of vehicles is constantly increasing. To proactively address future security risks, we present the Automotive Cybersecurity Trend Radar (ACTR), an approach that applies the Innovation Radar methodology to identify and temporally classify emerging technologies relevant for the automotive cybersecurity sector. By systematically analyzing technology trends, academic research, and industry reports, this radar provides a structured view of upcoming challenges and opportunities. Our research highlights key topics for the future, including the impact of post-quantum cryptography, AI-driven security mechanisms, and confidential computing technologies. By incorporating these insights into early-stage strategic planning, manufacturers and suppliers can improve their preparedness for emerging cybersecurity-relevant technologies as well as benefit from technology developments, allowing strategic investments. The ACTR thus serves as a tool for industry stakeholders to anticipate, prioritize, and address automotive cybersecurity challenges before they become critical.
Modern car insurance companies can dynamically adjust premiums with mobile sensor units that constantly record, analyze, and grade driving behavior. However, attackers can tamper with these systems to cheat the insurance company and reduce their premiums. We developed an attacker model for this threat in conjunction with a risk assessment using the ISO/SAE 21434 Risk Rating Methodology, a widely accepted standard for evaluating and managing risks in automotive systems. To support and showcase our model, we examined the static IT security of the mobile sensor unit DriveWell and its companion Android application HUK Mein Auto. In particular, we utilized digital forensic tools to analyze the stored data in the sensor’s internal storage and the app’s smartphone storage, with a focus on encryption mechanisms for locally stored data and internet communication. During this analysis, we spotted a critical vulnerability that enables attackers to manipulate their insurance scores and potentially reduce premiums by up to 30%. Our model and real-world analysis provide a solid foundation for reducing the risk of such manipulation of insurance premiums.
Trust is a fundamental requirement in vehicular systems, particularly in the context of autonomous and cooperative driving, where decisions often depend on data received from other vehicles or infrastructure. This paper presents a novel approach to integrating trust-based reasoning—specifically, decision-making based on subjective logic—into Attribute-Based Access Control (ABAC) systems, using the XACML policy language. We draw a parallel between trust management and access control, showing that trust assessments can be modeled as access decisions and expressed through standard access control policies. This integration allows for the reuse of mature ABAC infrastructure to represent and enforce trust-based decisions. Furthermore, we demonstrate how trust discounting mechanisms, which combine node-based and data-based trust, can be implemented within XACML policies. To evaluate this approach, we introduce a Rust-based XACML engine capable of interpreting trust-enhanced policies. Its integration in both inter- and intra-vehicle scenarios is discussed, assuming a modern zone-based architecture. Our results suggest that policy-based trust management can be efficiently implemented, enabling flexible and standardized trust reasoning in automotive environments.
Many automotive systems employ secure boot to protect the authenticity and integrity of its software. However, there are numerous ways that secure boot can be designed and implemented. For example, there is foreground and background secure boot as well as sequential, staged or hybrid secure boot. Secure boot is commonly achieved through the use of classical cryptographic algorithms such as RSA and ECC. With future advancements in quantum computing, it can be expected that these classical cryptographic algorithms will be broken and as such there is need for cryptographic agility to transition to post-quantum cryptography (PQC) algorithms. There are several NIST standardized PQC algorithms such as ML-KEM, HQC, ML-DSA, FN-DSA and SLH-DSA. In this paper, we explore the feasibility of using PQC algorithms for secure boot by implementing these PQC algorithms on a prototype system and performing measurements of the cryptographic operations. While there are many considerations for transitioning from classical cryptographic algorithms to PQC algorithms, this paper offers initial insights into which PQC algorithms may be useful for secure boot in automotive real-time systems.
Large Language Models (LLMs) are gaining traction in cybersecurity applications, offering both promising opportunities and potential new risks. The use of these models in sub-domains such as automotive is still in its early stages. In this work-in-progress study, we use GPT-4o from OpenAI to generate a preliminary set of domain-relevant cybersecurity questions exploiting the Automotive Information Sharing and Analysis Center (Auto-ISAC) framework, which we then refined through manual validation. We exploited the final set of 25 questions to evaluate the performance of five LLMs models. Then, these questions were administered through a survey to a group of 17 domain experts, allowing us to compare this baseline with the results from the LLMs. From our preliminary findings, we found that LLMs reached a mean of 91.2% of correct answers on the test while human experts’ performance reached 64.7%. This study lays the groundwork for future investigations into the use of LLMs in the automotive-security domain and into the safe and trustworthy exploitation of LLMs.
Testing cybersecurity properties of vehicle components is an inherently difficult task, especially if testing Electronic Control Units (ECUs) in isolation and outside of their intended operation environment. Rehosting ECU firmware is a challenging task because of internal and external runtime dependencies. Virtualization of the full system often fails to reach the required operational state because of unfulfilled preconditions in the distributed system environment. We propose NOTHORG, a framework architecture for creating digital twins of ECUs that allows emulation and runtime-stubbing of the required environment to facilitate dynamic cybersecurity testing. NOTHORG supports replaying external communication via Controller Area Network (CAN) bus and Ethernet and allows the integration of physical devices to the virtual environment. We validate our concept with a reference implementation and show that typical security testing tasks can be performed efficiently by combining static analysis, image preparation, dynamic runtime hooks, environment stubbing and log consolidation to reach operation states that would otherwise be unreachable in introspectable cybersecurity testing environments.
Quantum Key Distribution (QKD) enables two parties to generate a secret key, based on quantum physical properties. The secrecy of the exchanged key is guaranteed by these quantum physical properties, even against an attacker using a quantum computer. QKD therefore offers an Quantum Information Theoretically Secure (Q-ITS) alternative to Post Quantum Cryptography (PQC) schemes. This paper discusses advantages and challenges of deploying a QKD system in an Intelligent Transport System (ITS), by examining two exemplary use cases. Approaches to effectively implement QKD in these use cases are proposed, with a short discussion of advantages over the use of classical and PQC schemes.
Modern networked systems, such as those in the automotive sector, face increasing complexity and growing attack surfaces due to the rise of interconnected and data-driven technologies. Detecting malicious behavior in these environments requires efficient and scalable methods that can operate reliably despite limited resources and high communication volumes. This paper proposes a heuristic packet processing model designed to support intrusion detection based on structural and temporal characteristics of encrypted network traffic. The model follows a modular architecture consisting of four key phases: recording, sorting, prioritizing, and analyzing. At the core of the approach is the Polymetric Queueing Topology Space, a feature space that combines statistical and time series attributes derived from model structure and flow data. These features serve as input for machine learning models, which can effectively distinguish between benign and intrusion traffic patterns without relying on packet data beyond the transport layer. The approach was evaluated using the publicly available ToN_IoT dataset and demonstrated that reliable classification is achievable using a subset of the developed feature space that contains model-derived traffic features. We used Random Forest for supervised binary and multi-class classification achieving high accuracy scores of 99% for binary and 98% for multi-class classification. Additionally, for unsupervised anomaly detection, we created an Isolation Forest model accomplishing F1-scores of 0.92 for the benign and 0.96 for intrusion class. The architecture is designed to enable dynamic traffic prioritization and to offer a flexible foundation that can observe diverse network domains while maintaining efficient performance under constrained computational conditions.
The paper emphasizes some approaches to the problem of estimating either the transient time or the stability degree for linear and/or nonlinear systems. The methodologies taken into account rely on the root location for the characteristic equation and on the properties of the (strong) Lyapunov function(al)s.
Contextual emotion recognition systems often use environmental cues, such as social interactions and scene semantics, to infer emotional states. However, these cues can be unreliable for individuals with mental disorders due to atypical emotional expression and perception. This study investigates the efficacy of body pose as a robust alternative, given its direct correlation with physiological and affective states. We propose a two-stage methodology that involves pose detection, depth estimation, and a fusion model that integrates positional data. Evaluated on the BEAST dataset, our approach demonstrates a 35% reduction in misclassification rates compared to state-of-the-art context-aware models, particularly for bipolarity, depression, and anxiety populations. These findings suggest a need to reconsider emotion recognition pipelines for clinical and assistive technologies, prioritizing body pose as a primary input.
This review aims to identify the most significant practical applications, limitations, and future directions of generative AI (artificial intelligence) and LLMs (Large Language Models) to guide researchers, industry stakeholders, and policymakers. Through comprehensive analysis of scientific literature, it was traced the development of these technologies from early linguistic theories to modern transformer-based architectures. The findings presented in this review article reveal the transformative impact of LLMs across diverse domains including healthcare, education, software development, and creative industries. Significant technical limitations were identified, including hallucinations, context window constraints, and reasoning deficiencies, along-side ethical concerns regarding bias, privacy, and environmental impact. The review concludes by exploring emerging trends in model architecture, efficiency improvements, and ethical frameworks that will shape future development. This work provides researchers, practitioners, and policymakers with a comprehensive understanding of the current state and future trajectory of generative AI and LLMs.
This study assesses the emerging practice of generative content analysis through a comparative evaluation of how five generative AI models (ChatGPT-4o, ChatGPT-4.5, Claude-3.7 Sonnet, Gemini-2.0 Flash, and Zeno) perform thematic analysis on qualitative accounts of "love at first sight" experiences. Drawing on Flaherty's concept of "time work", referring to attempts to customize temporal experience, we examine how these models analyze temporality in romantic narratives. Using an analytical framework focusing on five temporal themes, we compare the models' convergence, divergence, and complementarity in their interpretations when analyzing a discussion thread on the social platform Reddit. Results reveal strong consistency at the theme level but significant variation in subtheme identification and quote selection, with most quotes (58 of 65) uniquely identified by only one model. When analyzing identical quotes, models agreed on theme classification in only about half of instances. The analysis demonstrates complementarity between models, with optimal combinations strongly improving analytical coverage. Beyond methodological findings, the substantive results uncover how love at first sight experiences function as a form of time work and resistance to rationalized temporality in modern society, creating subjective "temporal bubbles" outside institutional time structures. Temporal distortion phenomena, immediate felt familiarity, and accelerated relationship progression serve as legitimizing frameworks for relationship decisions that depart from conventional norms. This research contributes to understanding both the potential of AI-assisted qualitative analysis and how romantic temporal experiences represent distinctive forms of temporal agency.
Electromyography (EMG) signals play a crucial role in diagnosing neuromuscular disorders by distinguishing between normal and pathological muscle activity. This study explores a machine learning-based classification approach to identify abnormalities in EMG signals. A dataset consisting of signal recordings represented as images is used. They are processed through feature extraction and normalization techniques to improve classification performance. The experimental results highlight the potential of machine learning models in automating EMG signal classification, offering valuable insight for clinical diagnostics and rehabilitation applications.
Innovative agricultural technology provides international coverage for practical applications and intelligent systems integrating advanced computing. Among those technologies introduced were artificial intelligence (AI) and machine learning. Thus, ongoing research focuses on improving CNN architectures and exploring their applications across various fields, including agriculture domain. However, in the agricultural domain, specifically for classifying Tikog leaves, a raw material used in the banig weaving industry utilizing feature engineering and optimization of Convolutional Neural Network (CNN) using featuring engineering approach remains unexplored. Therefore, this study investigated the application of CNNs in the agricultural domain, specifically for classifying Tikog, a raw material used in the banig weaving industry. This study utilized the KDD process, employing various machine learning algorithms such as color lab space representation, data augmentation, feature extraction, pooling technique, and fine-tuning of 500 Standard Quality and 500 Substandard Quality original data and augmented to 3000 images and 500 images for testing or validation purposes. Results revealed that there were 27 cases optimized models generated, and four (4) emerged as top-performing models, with Case 12 as the best-optimized model with accuracy (94.23%, 96.83%), precision (92.45%, 95.77%), recall (96.33%, 98.00%), F1-score (94.35%, 96.87%), ROC/AUC (98.18%, 99.40%) and SSE (173, 19). Case 12 optimized CNN model shows a non-overfitting model with computational efficiency and is highly effective, reinforcing its suitability as the best-performing model among the cases.
Recognizing human activities purely from movement patterns has many applications in healthcare and fitness tracking. Although machine learning models that achieve high accuracy have been developed, their use in resource-constrained embedded devices remains challenging. This paper reviews the state of the art in Human Activity Recognition (HAR) and presents a CNN designed to detect walking trained on a large, publicly available, free-living dataset. The model is also used together with an open source step counting algorithm in order to measure its accuracy on unseen data.
Autonomous vehicles represent a subject of worldwide importance, particularly in academic and industrial research. This field efficiently integrates theoretical concepts of modeling and control with practical applications, yielding innovation in the development and implementation of new ideas and concepts. This study presents a novel approach to enhance vehicle dynamics control through the application of nonlinear model predictive control (NMPC) with a focus on the essential nonlinearities that can influence vehicle dynamics. The proposed method offers significant advantages over traditional linear control techniques, particularly in scenarios in which the vehicle’s behavior deviates from linearity, such as difficult maneuvers, employing techniques for the implementation of the trajectory tracking methods and the generation of the velocity profile. Furthermore, this paper opens new potential perspectives in real-world applications and avenues for future research, pointing out the usage of NMPC in advancing vehicle dynamic control systems and integration of smart functionalities.
Hyperbaric Oxygen Therapy surgical rooms require precise pressure control to ensure patient safety and therapeutic efficacy. This paper presents a comprehensive study on advanced control strategies for pressure regulation in hyperbaric environments. A dynamic process model of the hyperbaric chamber is developed, based on thermodynamic and fluid dynamics principles, and implemented in MATLAB/Simulink to simulate real-world operational scenarios. Three distinct control strategies are proposed and evaluated: a conventional Proportional-Integral-Derivative controller, a Fractional-Order PID controller for enhanced nonlinear system handling, and an Adaptive FOPID controller integrated with artificial neural networks to dynamically optimize control parameters in real time. Simulation results demonstrate that the Adaptive FOPID controller outperforms its counterparts, achieving superior transient response and reduced settling time. The integration of AI-driven neural networks enables adaptive tuning of fractional-order parameters, ensuring optimal performance across diverse pressure profiles. This work highlights the potential of adaptive fractional-order control systems in critical healthcare environments, offering a framework for safe, efficient, and reliable hyperbaric oxygen therapy.
