Kyriacos E. Pavlou
Research
The DiamondTouch Project
This first research project started as part of my CSc 445: Algorithms class with Dr. Stephen Kobourov. What began vaguely as a graph-related project, evolved into an HCI study with a marked statistical bias. The initial premise was to investigate the efficiency of a new input device, when used to solve graph problems, and determine whether it could serve as a better alternative to the ubiquitous mouse-monitor setup. We focused on the DiamondTouch table, a touch-sensitive input device developed at the Mitsubishi Electric Research Laboratories. We proceeded to compare the efficiency of the traditional mouse-monitor interface against that of the DiamondTouch table when solving spatially oriented problems and when completing such tasks collaboratively. We timed how long it took a group of individuals and a group of collaborative pairs to solve the problem of untangling planar graphs using the traditional setup versus the DiamondTouch table.
Preliminary findings indicate that there is no significant difference between the mean time required to solve the graph problems when individual users work with the mouse-monitor setup and when they use the DiamondTouch hardware. Thus, we can infer that the DiamondTouch table does not confer any advantage to its users over the individuals who employed the mouse-monitor setup to solve the problems. In the case of paired users, however, the data suggest that there is a 95% probability for a significant difference to exist between the means of the two populations. Pairs using the DiamondTouch table were able to solve the graph problems faster than the ones using the mouse-monitor setup, an indication that the DiamondTouch table confers an advantage to collaborative users
We proceeded to add an extra layer of complexity, namely, to perform a further test on group users in which two different people use two different mice on a single machine to manipulate the nodes of a single graph. Even with this scenario the DiamondTouch pairs continued to outperform the users in all the other experimental configurations. This strengthens the validity of our results by eliminating the possibility the observed difference in mean times of completion is due to division of labor.
Our results were published at INTERACT 2005, the Tenth IFIP TC13 International Conference on Human-Computer Interaction.
The τZaman Project
This research project was undertaken as an independent study (CSc 499) with Dr. Richard Snodgrass. The project is part of the τZaman system which provides a sophisticated Java API for the manipulation of time-oriented data.
My contribution to this project was to realize a BNF grammar and a denotational semantics for an XML specification of calendars. This will be used to implement a calendric algebra allowing direct manipulation and conversions between time granularities. The advantage of this approach is that it allows a direct conversion of a time value in a granularity, specified as algebraic operations of arbitrary nesting depth, to any other granularity without having to convert the initial granularity to a base granularity as an intermediate step. The denotational semantics will then be used to implement a compiler that will take as input XML documents defining the original granularities. The compiler will parse the specification and produce two output files: a calendar specification file describing the definition of the new converted granularity and an output file with the Java code used to perform the conversion.
The algebraic operators implemented are based on the algebra developed by Ning, Wang, and Jajodia, which appeared in the Annals of Mathematics and Artificial Intelligence. I am still working on this project, and when the project is complete, we intend to submit a technical report to IEEE Transactions on Knowledge and Data Engineering.
Audit Log Security
This research is a natural continuation of the previously published work done on tamper detection in audit logs [R. T. Snodgrass, et al. "Tamper Detection in Audit Logs," VLDB, pp. 504-515, Toronto, Canada, September 2004]. The results of this current work were published at SIGMOD 2006. The abstract of the paper is given below.
Abstract: Mechanisms now exist that detect tampering of a database, through the use of cryptographically-strong hash functions. This paper addresses the next problem, that of determining who, when, and what, by providing a systematic means of performing forensic analysis after such tampering has been uncovered. We introduce a schematic representation termed a "corruption diagram" that aids in intrusion investigation. We use these diagrams to fully analyze the original proposal, that of a linked sequence of hash values. We examine the various kinds of intrusions that are possible, including retroactive, introactive, backdating, and postdating intrusions. We then introduce successively more sophisticated forensic analysis algorithms: the monochromatic, RGB, and polychromatic algorithms, and characterize the "forensic strength" of these algorithms. We show how forensic analysis can efficiently extract a good deal of information concerning a corruption event.
Genome Alignment
in progress...