Supporting model based safety and security assessment of high assurance systems

Abstract

Modern embedded systems are more complex than ever due to intricate interaction with the physical world in a system environment and sophisticated software in a resource-constrained context. Cyber attacks in software-reliant and networked safety-critical systems lead to consideration of security aspects from the system’s inception. Model-Based Development (MBD) is one approach that has been an effective development practice because of the abstraction mechanism that hides the complicated lower-level details of software and hardware components. Standards play an essential role in embedded development to ensure the safety of the users and environment. In safety-critical domains like avionics, automotive, and medical devices, standards provide best practices and consistent approaches across the community. The Analysis and Design Language (AADL) is a standardized modeling language that includes patterns that reflect best architectural practices inspired by multiple safety-critical domains. The work described in this dissertation comprises numerous contributions that support a model analysis framework for AADL that aims to help developers design and assure safety and security requirements and demonstrate system conformance to specific categories of standards. This first contribution is Awas - an open-source framework for performing reachability analysis on AADL models annotated with information flow annotations at varying degrees of detail. The framework provides highly scalable interactive visualizations of flows with dynamic querying capabilities. Awas provide a simple domain-specific language to ease posing various queries to check information flow properties in the model. The second contribution is a process for integrating risk management tasks of ISO 14971 - the primary risk management standard in the medical device domain — with AADL modeling, specifically with AADL’s error modeling (EM) of fault and error propagations. This work uses an open-source patient-controlled analgesic (PCA) pump - the largest open-source AADL model to illustrate the integration of risk management process with AADL and provides the first mapping of AADL EM to ISO 14971 concepts. It also provides industry engineers, academic researchers, and regulators with a complex example that can be used to investigate methodologies and methods of integrating MBD and risk management. The third contribution is a technique to model and analyze security properties such as confidentiality, authentication, and resource partitioning within AADL models. This effort comprises an AADL annex language to model multi-level security domains along with classification of system elements and data using those domains and a tool to infer security levels and check information leaks. The annex language and the tools are evaluated and integrated into the AADL development environment for a seamless workflow.