Use this URL to cite or link to this record in EThOS:
Title: Automata based monitoring and mining of execution traces
Author: Reger, Giles Matthew
ISNI:       0000 0004 5352 9629
Awarding Body: University of Manchester
Current Institution: University of Manchester
Date of Award: 2014
Availability of Full Text:
Access from EThOS:
Access from Institution:
This thesis contributes work to the fields of runtime monitoring and specification mining. It develops a formalism for specifying patterns of behaviour in execution traces and defines techniques for checking these patterns in, and extracting patterns from, traces. These techniques represent an extension in the expressiveness of properties that can be efficiently and effectively monitored and mined. The behaviour of a computer system is considered in terms of the actions it performs, captured in execution traces. Patterns of behaviour, formally defined in trace specifications, denote the traces that the system should (or should not) exhibit. The main task this work considers is that of checking that the system conforms to the specification i.e. is correct. Additionally, trace specifications can be used to document behaviour to aid maintenance and development. However, formal specifications are often missing or incomplete, hence the mining activity. Previous work in the field of runtime monitoring (checking execution traces) has tended to either focus on efficiency or expressiveness, with different approaches making different trade-offs. This work considers both, achieving the expressiveness of the most expressive existing tools whilst remaining competitive with the most efficient. These elements of expressiveness and efficiency depend on the specification formalism used. Therefore, we introduce quantified event automata for describing patterns of behaviour in execution traces and then develop a range of efficient monitoring algorithms. To monitor execution traces we need a formal description of expected behaviour. However, these are often difficult to write - especially as there is often a lack of understanding of actual behaviour. The field of specification mining aims to explain the behaviour present in execution traces by extracting specifications that conform to those traces. Previous work in this area has primarily been limited to simple specifications that do not consider data. By leveraging the quantified event automata formalism, and its efficient trace checking procedures, we introduce a generate-and-check style mining framework capable of accurately extracting complex specifications. This thesis, therefore, makes separate significant contributions to the fields of runtime monitoring and specification mining. This work generalises and extends existing techniques in runtime monitoring, enabling future research to better understand the interaction between expressiveness and efficiency. This work combines and extends previous approaches to specification mining, increasing the expressiveness of specifications that can be mined.
Supervisor: Rydeheard, David Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: Runtime monitoring ; Runtime verification ; Specification mining ; Trace analysis