UC Irvine

Smart systems are software entities that carry out a set of operations on behalf of a user or another application with some degree of independence or autonomy. These systems employ some knowledge or representations of (1) a user's goals or desires, and (2) the environment in which they act in to achieve these goals. Such an agent is a system situated in a technical or natural environment that senses some status from that environment and acts on it, changing part of its environment or influencing what it senses.

The main characteristics of smart systems are adaptive capacity, indicating that such systems adapt as information changes, and they may resolve ambiguity and tolerate unpredictability; learning capability, implying that smart systems reason on data to create new information and use closed-loop feedback to learn from the output; context awareness, indicating that smart systems may identify and extract contextual elements such as syntax, time, location, etc; and dynamic interactivity, indicating that these systems can interact with users or other applications and cloud services to understand their goals and needs.

Smart systems address environmental, societal, and economic challenges like limited resources, climate change, and globalization. They are, for that reason, increasingly used in a large number of sectors such as transportation, healthcare, energy, safety, security, etc. Hence, the need for effective analysis and testing techniques for such systems has increased more than ever.

This dissertation proposes to ensure the dependability (i.e., security, safety, and reliability) of smart systems by (1) analyzing bugs and vulnerabilities found in such systems and (2) developing tools to test and detect bugs and vulnerabilities in smart systems automatically.

Experiments conducted on real-world, open-source software applications corroborates the effectiveness and efficiency of our proposed approaches and their ability to ensure the dependability of software in smart systems.