Integrating Artificial Intelligence (AI) technology in electric vehicles (EV) introduces unique challenges for safety assurance, particularly within the framework of ISO 26262, which governs functional safety in the automotive domain. Traditional assessment methodologies are not geared toward evaluating AI-based functions and require evolving standards and practices. This paper explores how an independent assessment of an AI component in an EV can be achieved when combining ISO 26262 with the recently released ISO/PAS 8800, whose scope is AI safety for road vehicles. The AI-driven State of Charge (SOC) battery estimation exemplifies the process. Key features relevant to the independent assessment of this extended evaluation approach are identified. As part of the evaluation, robustness testing of the AI component is conducted using fault injection experiments, wherein perturbed sensor inputs are systematically introduced to assess the component's resilience to input variance.

Fault- and attack injection are techniques usedto measure dependability attributes of  omputer systems. Animportant property of such techniques is their efficiency inexploring the target system’s fault- or attack space. As this spaceis generally very large, pre-injection analysis techniques may beused to effectively explore the space. In this paper, we studytwo such techniques proposed in the past, namely inject-on-readand inject-on-write. Furthermore, we propose two new techniquescalled error space pruning of signals and error space pruning ofsignals and ports and evaluate their efficiency in reducing thespace needed to be explored by injection experiments. Thesetechniques were integrated into MODIFI, a fault- and attackinjector targeting Simulink models. To the best of our knowledge,we are the first to evaluate these pre-injection techniques for thiskind of injector.The results of our evaluation of 11 Simulink models from theautomotive domain and one from the avionics domain, show thatthe new proposed techniques reduce the fault- and attack spaceneeded to be explored by about 27–49%. Using MODIFI, we thenperformed injection experiments on two automotive models,  swell as an aero engine control model, while elaborating on theresults obtained.

he increased importance of cybersecurity in autonomous machinery is becoming evident in the forestry domain. Forestry worksites are becoming more complex with the involvement of multiple systems and system of systems. Hence, there is a need to investigate how to address cybersecurity challenges for autonomous systems of systems in the forestry domain. Using a literature review and adapting standards from similar domains, as well as collaborative sessions with domain experts, we identify challenges towards CE-certified autonomous forestry machines focusing on cybersecurity and safety. Furthermore, we discuss the relationship between safety and cybersecurity risk assessment and their relation to AI, highlighting the need for a holistic methodology for their assurance.

A remotely operated road vehicle (RORV) refers to a vehicle operated wirelessly from a remote location. In this paper, we report results from an evaluation of two safety mechanisms: safe braking and disconnection. These safety mechanisms are included in the control software for RORV developed by Roboauto, an intelligent mobility solutions provider. The safety mechanisms monitor the communication system to detect packet transmission delays, lost messages, and outages caused by naturally occurring interference as well as denial-of-service (DoS) attacks. When the delay in the communication channel exceeds certain threshold values, the safety mechanisms are to initiate control actions to reduce the vehicle speed or stop the affected vehicle safely as soon as possible. To evaluate the effectiveness of the safety mechanisms, we exposed the vehicle control software to various communication failures using a software-in-the-loop (SIL) testing environment developed specifically for this study. Our results show that the safety mechanisms behaved correctly for a vast majority of the simulated communication failures. However, in a few cases, we noted that the safety mechanisms were triggered incorrectly, either too early or too late, according to the system specification. 

A remotely operated road vehicle (RORV) refers to a vehicle operated wirelessly from a remote location. In this paper, we report results from an evaluation of two safety mechanisms: safe braking and disconnection. These safety mechanisms are included in the control software for RORV developed by Roboauto, an intelligent mobility solutions provider. The safety mechanisms monitor the communication system to detect packet transmission delays, lost messages, and outages caused by naturally occurring interference as well as denial-of-service (DoS) attacks. When the delay in the communication channel exceeds certain threshold values, the safety mechanisms are to initiate control actions to reduce the vehicle speed or stop the affected vehicle safely as soon as possible. To evaluate the effectiveness of the safety mechanisms, we exposed the vehicle control software to various communication failures using a software-in-the-loop (SIL) testing environment developed specifically for this study. Our results show that the safety mechanisms behaved correctly for a vast majority of the simulated communication failures. However, in a few cases, we noted that the safety mechanisms were triggered incorrectly, either too early or too late, according to the system specification.

As vehicles become more and more connected with their surroundings and utilize an increasing number of services, they also become more exposed to threats as the attack surface increases. With increasing attack surfaces and continuing challenges of eliminating vulnerabilities, vehicles need to be designed to work even under malicious activities, i.e., under attacks. In this paper, we present a resilience framework that integrates analysis of safety and cybersecurity mechanisms. We also integrate resilience for safety and cybersecurity into the fault – error – failure chain. The framework is useful for analyzing the propagation of faults and attacks between different system layers. This facilitates identification of adequate resilience mechanisms at different system layers as well as deriving suitable test cases for verification and validation of system resilience using fault and attack injection.

As more parts of the power grid become connected to the internet, the risk of cyberattacks increases. To identify the cybersecurity threats and subsequently reduce vulnerabilities, the common practice is to carry out a cybersecurity risk assessment. For safety classified systems and products, there is also a need for safety risk assessments in addition to the cybersecurity risk assessment to identify and reduce safety risks. These two risk assessments are usually done separately, but since cybersecurity and functional safety are often related, a more comprehensive method covering both aspects is needed. Some work addressing this has been done for specific domains like the automotive domain, but more general methods suitable for, e.g., Intelligent Distributed Grids, are still missing. One such method from the automotive domain is the Security-Aware Hazard Analysis and Risk Assessment (SAHARA) method that combines safety and cybersecurity risk assessments. This paper presents an approach where the SAHARA method has been modified to be more suitable for larger distributed systems. The adapted SAHARA method has a more general risk assessment approach than the original SAHARA. The proposed method has been successfully applied on two use cases of an intelligent distributed grid.

As more parts of the power grid become connected to the internet, the risk of cyberattacks increases. To identify the cybersecurity threats and subsequently reduce vulnerabilities, the common practice is to carry out a cybersecurity risk assessment. For safety classified systems and products, there is also a need for safety risk assessments in addition to the cybersecurity risk assessment in order to identify and reduce safety risks. These two risk assessments are usually done separately, but since cybersecurity and functional safety are often related, a more comprehensive method covering both aspects is needed. Some work addressing this has been done for specific domains like the automotive domain, but more general methods suitable for, e.g., Intelligent Distributed Grids, are still missing. One such method from the automotive domain is the Security-Aware Hazard Analysis and Risk Assessment (SAHARA) method that combines safety and cybersecurity risk assessments. This paper presents an approach where the SAHARA method has been modified in order to be more suitable for larger distributed systems. The adapted SAHARA method has a more general risk assessment approach than the original SAHARA. The proposed method has been successfully applied on two use cases of an intelligent distributed grid.

This paper presents ComFASE, a communication fault and attack simulation engine. ComFASE is used to identify and evaluate potentially dangerous behaviours of interconnected automated vehicles in the presence of faults and attacks in wireless vehicular networks. ComFASE is built on top of OMNET++ (a network simulator) and integrates SUMO (a traffic simulator) and Veins (a vehicular network simulator). The tool is flexible in modelling different types of faults and attacks and can be effectively used to study the interplay between safety and cybersecurity attributes by injecting cybersecurity attacks and evaluating their safety implications. To demonstrate the tool, we present results from a series of simulation experiments, where we injected delay and denial-of-service attacks on wireless messages exchanged between vehicles in a platooning application. The results show how different variants of attacks influence the platooning system in terms of collision incidents.

As society increasingly relies on safety- and security- critical systems, the need for confirming their dependability becomes essential. Adequate V&V (verification and validation) methods must be employed, e.g., for system testing. When selecting and using the methods, it is important to analyze their possible gaps and limitations, such as scalability issues. However, and as we have experienced, common, explicitly defined criteria are seldom used for such analyses. This results in analyses that consider different aspects and to a different extent, hindering their comparison and thus the comparison of the V&V methods. As a solution, we present a set of criteria for the analysis of gaps and limitations of V&V methods for safety- and security-critical systems. The criteria have been identified in the scope of the VALU3S project. Sixty-two people from 33 organizations agreed upon the use of nine criteria: functionality, accuracy, scalability, deployment, learning curve, automation, reference environment, cost, and standards. Their use led to more homogeneous and more detailed analyses when compared to similar previous efforts. We argue that the proposed criteria can be helpful to others when having to deal with similar activities.