Towards Automated Threat Modeling of Cyber-Physical Systems

Abstract

Cyber-Physical System (CPS) seamlessly integrates the computation, communication, and physical components of the system. Often, a CPS controls physical objects through computation and communication and uses of real-time feedback. Threat models of such systems must consider their hardware, network, infrastructure, software, and human aspects and the interactions of these aspects. Commonly, threat modeling of such systems is based on the given system's architecture. In terms of components and interactions among these components, the architecture of a given CPS may change over time, making the threat model of the CPS rapidly obsolete-i.e., incomplete and invalid threat model. This paper poses the question: Can we automate threat modeling of a given CPS? A positive answer to the question helps to implement continuous up-to-date security assessments of CPSs-for different versions of the given system. It presents an approach to maintain the threat model of given CPSs up-to-date and reports about applying the proposed approach on Apollo Auto 3.5, an autonomous vehicle software. Unfortunately, the scalability limitation of the used architecture recovery technique prevents the recovering the Apollo Auto architecture and, consequently, the automated identification of the system's threat model.

Challenges in threat modelling of new space systems: A teleoperation use-case

“…These systems include distributed units with some degree of autonomous functions. However, it is difficult to cover all aspects of such use cases by using a single method Jamil et al (2021a). Thus, the methods are evaluated for the selection in our study, based on the metrics derived from the work in Shevchenko et al (2018).…”

Securing Cloud-Assisted Connected and Autonomous Vehicles: An In-Depth Threat Analysis and Risk Assessment

“…We considered five potentially relevant threat modeling methods for a CCAV ecosystem [ 19 ]. It is difficult to address all the challenges of such a use case with a single solution [ 20 ]. Therefore, a selection of metrics developed from the work in [ 21 ] were used to assess the methods in this research: (i) Maturity: Is the technique well-defined and has it been employed in earlier research?…”

Data Privacy Threat Modelling for Autonomous Systems: A Survey From the GDPR's Perspective

“…The UAV has a complex architecture and most of the information regarding measures in place is confidential, so it is difficult to identify which threat is more dangerous to a UAV that needs to be modelled at utmost priority [106]. Moreover, there has been wide research for modelling the security and privacy of vehicleto-vehicle (V2V) [117], Vehicle-to-infrastructure (V2I) [118] and Vehicle-to-everything (V2X) [119]. But threat modelling techniques (such as STRIDE, LINDDUN) are not sufficient for securing V2V, V2I and V2X because they require the autonomy approach [120] to deal with privacy and security vulnerabilities.…”