Fault-based Testing of Evolving Real-time Systems

点击翻译按钮获取中文摘要

Failures of evolving real-time systems are not seldom. These systems include but are not limited to control and security systems, communication protocols in domains such as the aerospace, the automotive, finance and the popular web and mobile applications. A real-time system maintains a current state that it uses to react to the stimuli from the environment and it is organized in communicating components. Components can evolve following the end-user needs or change in the environment. Failures of the systems can be caused by logical and timing faults, or vulnerabilities introduced at the different phases of the system design. Failures result in damages yielding to the loss of confidence in the systems and their rejection by the end-users. The long-term objective of this research program is to propose and develop a fault-based testing theory that can be applied to detect logical, timing and security faults in evolving real-time systems. We will consider faults introduced at the unit and integration phases of the system development. I plan to use formal representations (specifications) of logical, timing and security properties of systems such as state transition machines. This work will provide testing tool developers with efficient methods for the fault detection in evolving real-time systems. The Novelty of the proposed research includes proposing an approach to build specifications of evolving real-time systems and a fault model to represent specific faults to be detected, and methods of test analysis, test generation and test selection for the detection of the specific unit and integration faults. A fault model is defined as a tuple of a specification, a fault domain and a conformance relation. The fault domain represents a set of faulty implementations obtained by seeding faults in the specification. Specific faults can be made available in databases or by an expert; a limited number of tests allows detecting them with reduced testing efforts, which is in line with the objectives and practices of the industry. We also generate optimal tests having reduced execution times. We will use deterministic or nondeterministic models with timing and probabilistic constraints to capture logical, timing and evolution properties of the systems. The methods will benefit from the recent advances in fields such as constraint solving, symbolic execution, abstract interpretation and model learning. We will implement the proposed methods in proof-of-concept tools. The expected results will have a significant impact on the scientific knowledge and industrial practices in testing. Numerous industrial projects executed by CRIM create a proper opportunity to increase awareness of industrial partners about systematic approaches in testing. Ten HQPs (2 PhD, 3 MSc, 5 BSc) will participate in this research. The skills gained by the HQPs in formal modeling and fault-based testing will be beneficial to all of them and to the development evolving real-time systems.