Security & Trust in Multi-Agent Systems
Multi-agent systems (MAS) consist of a collection of agents
(software programs) that interact with each other in dynamic,
unpredictable environments. A fundamental question in MAS is whether
messages, sent and received by agents, have been regarded as trustworthy.
Therefore, it is important to provide a formal method for specifying trust
that agents place in the security mechanisms of the system,
so as to support reasoning about its security properties.
Various formalisms have been proposed for specifying trust mechanisms in MAS.
Most of them only deal with the static aspects of trust and ignore
the evolution of trust relationships over time.
This project is developing a novel formalism for specifying dynamic
trust relationships as the agents in MAS gain or lose confidence
in the agents and the environment in which they operate.
Certain aspects of this project has been funded by MURDG's and a
multi-campus ARC Discovery Project (2005-2007) held at
the University of Western Sydney.
Formal Approaches to Intelligent Systems
Logical representations have been widely used in Computer
Science and Artificial Intelligence (AI).
However, as the problems tackled have become more complex,
the requirement for more powerful logical representations has been growing.
In particular, since the concept of time is of central importance
to an increasingly wide range of applications, including
the representation of time-dependent data, modeling reactive systems,
and the specification and verification of concurrent and distributed systems,
many logics incorporating temporal notions are being developed and applied.
Then the development of executable temporal logics followed so that
system developers have access to these, more powerful, logical techniques.
We have proposed and investigated one of the earliest executable
temporal logic systems called Chronolog (for Chronological logic programming).
Chronolog is a language in which one can specify dynamic systems
whose behaviour varies with time.
We have also investigated extensions of Chronolog where one can specify
and reason about the time-dependent properties of reactive, multi-agent
systems where components (and agents) run on different clocks with
varying rates of progress.
Dr. C. Liu from DSTO (Edinburgh) and Prof. W. Wadge from
University of Victoria (Canada) have been involved in the theoretical
aspects of the project.
The implementation work on Chronolog has been carried out
jointly with Prof. K. Zhang of University of Texas, Dallas.
Aspects of this project has been funded by MURGs, small ARC grants,
one ARC Discovery Project under Innovation Initiative(2002),
and currently by a 5-year, multi-campus ARC Discovery Project (2004-2008)
held at Griffith University, involving five investigators from
five different institutions.
Temporal Models and Algebras for Databases
The ability to model the temporal dimension of real world is essential
to many computer system applications, such as banking, inventory control,
student records, and airline reservations. Yet conventional database
systems do not adequately support the time-varying aspect of
the real-world phenomena. This project was concerned with
the investigation of the fundamental problems in representing and
querying time-varying information with the relational database technology,
and providing application-independent solutions to those problems.
The focus was on incorporating an implicit time dimension
into the relational model and algebra, through the use of temporal logic.
The resulting temporal algebra can be used as a testbed for the further study
and exploration of the important properties such as completeness
and the expressive capabilities that temporal query languages should exhibit.
We also showed that a representation-independent temporal data model
based on temporal logic makes possible the study of integration and
interoperability of different, and representation-dependent
temporal data models within a single, precise framework.
This project was mainly funded by a large ARC grant (1996-1998).
Modelling and Reasoning about Multidimensional Information
In many computer applications, multidimensional information is pervasive,
for example, time series, forex data, spatial information, weather
forecast data and so on. Multidimensional nature of such applications
also demands more sophisticated techniques and tools for the representation
and manipulation of multidimensional information. Most of the approaches
proposed so far have been application-dependent, they do not have a clear
and well-defined underlying formalism, and hence they do not easily
generalise to other application domains.
These shortcomings restrict their wide acceptance and use.
We demonstrated that the use of executable multidimensional logics enhanced
with aggregation operators in support of OLAP functionalities
can remedy some of the shortcomings of the existing approaches.
I have collaborated with Dr. R. Wong of UNSW/NICTA and A/Prof. W. Du of
The University of New Brunswick (Canada) in this project which
was mainly funded by MURGs and a large ARC grant (2000-2002).
Temporal Data Mining
In many business applications, a great deal of temporal data has
been stored in corporate data warehouses. Examples range from
transaction databases in health care and insurance, patient records,
and stock exchange, to scientific databases in geophysics and astronomy.
The growth of research and applications on temporal databases and
data warehouses has motivated an urgent need for techniques to deal with
the problem of knowledge discovery from large temporal databases.
This project is concerned with the investigation of a model for temporal
support in data mining and associated techniques and methods.
This project has been funded by MURGs,
an ongoing ARC Linkage grant with the Australian Taxation Office (ATO) as
the industry partner, and more recently Capital Markets CRC project seed funding.
Software reverse engineering
This project was concerned with the development of tools and techniques for
reverse engineering and software analysis. Reverse engineering is
the process of extracting system abstractions and design information
out of existing software systems for maintenance and re-engineering purposes.
This process involves the identification of software artifacts
in a subject system, the exploration of the way in which these artifacts
interact with one another, and the aggregation of the artifacts to form
more abstract system representations that facilitate program understanding.
Of particular interest was the Rigi approach to reverse engineering based
on module interconnection models and subsystem hierarchies identified
by interactive, semi-automatic reconstruction of system representations.
This was a joint project with Prof. Hausi Muller of the
University of Victoria (Canada), primarily funded by MURGs.