CONSIDERING SAFETY AND SECURITY IN AV FUNCTIONS

Abstract

Autonomous vehicles (AVs) are coming to our streets. Due to the presence of highly complex software systems in AVs, a new hazard analysis technique is needed to meet stringent safety standards. Also, safety and security are inter-dependent and inter-related aspects of AV. They are focused on shielding the vehicles from deliberate attacks (security issue) as well as accidental failures (safety concern), that might lead to loss of lives and injuries to the occupants. So, the current research work has two key components: functional safety and cybersecurity of the autonomous systems.

For the safety analysis, we have applied System Theoretic Process Analysis (STPA), which is built on Systems Theoretic Accident Modeling and Processes (STAMP). STAMP is a powerful tool that can identify, define, analyze, and mitigate hazards from the earliest conceptual stage of development to the operation of a system. Applying STPA to autonomous vehicles demonstrates STPA's applicability to preliminary hazard analysis, alternative available, developmental tests, organizational design, and functional design of each unique safety operation.

This thesis describes the STPA process used to generate system design requirements for an Autonomous Emergency Braking (AEB) system using a top-down analysis approach for the system safety. The research makes the following contributions to practicing STPA for safety and security:

1. It describes the incorporation of safety and security analysis in one process and discusses the benefits of this; 2. It provides an improved, structural approach for scenario analysis, concentrating on safety and security; 3. It demonstrates the utility of STPA for gap analysis of existing designs in the automotive domain; 4. It provides lessons learned throughout the process of applying STPA and STPA-Sec.

Controlling a physical process is associated with dependability requirements in a cyber-physical system (CPS). Cyberattacks can lead to the dependability requirements not being in the acceptable range. Thus, monitoring of the cyber-physical system becomes inevitable for the detection of the deviations in the system from normal operation. One of the main issues is understanding the rationale behind these variations in a reliable manner.

Understanding the reason for the variation is crucial in the execution of accurate and time-based control resolution, for mitigating the cyberattacks as well as other reasons of reduced dependability. Currently, we are using evidential networks to solve the reliability issue. In the present work, we are presenting a cyber-physical system analysis where the evidential networks are used for the detection of attacks.

The results obtained from the STPA analysis, which provides the technical safety requirements, can be combined with the EN analysis, which can be used efficiently to detect the quality of the used sensor to justify whether the CPS is suitable for the safe and secure design.