BUKU ABSTRAK - Universiti Putra Malaysia

Anonymous Author Discovery from Writing Style
Dr. Norwati Mustapha
Tareef Kamil Mustafa, Masrah Azrifah Azmi Murad and Md. Nasir Sulaiman
Faculty of Computer Science and Information Technology, University Putra Malaysia,
43400 UPM Serdang, Selangor, Malaysia.
+603-8946 6585; norwati@fsktm.upm.edu.my
Keywords: Text mining, stylometric, authorship attribution
UPEM: User-centred Privacy Evaluation Model in Pervasive Computing Systems
Dr. Nur Izura Udzir
Ali Dehghantanha, Ramlan Mahmod and Zuriati Ahmad Zukarnain
Faculty of Computer Science and Information Technology, University Putra Malaysia,
43400 UPM Serdang, Selangor, Malaysia.
+603-8946 6521; izura@fsktm.upm.edu.my
Keywords: Privacy, pervasive computing, ubiquitous computing, privacy evaluation
217
Science, Technology & Engineering
Stylometric Authorship attribution is one of the new approaches in the text mining field that has recently been
noticed because of its delicateness. This approach is concerned about analysing texts, e.g. novels and plays that
famous authors wrote, trying to measure the authors’ style, by choosing some attributes that shows the authors’
style of writing, assuming that these writers have a special way of writing, that no other writer has. The two major
problems that should be solved in this field are predictions accuracy results and human judgment dependency
(i.e. expert). The techniques that manage these kinds of predictions are either using the computational statistical
frequent word or the use of more sophisticated semantic techniques but they are still not completely accurate.
In this research, we propose a new Stylometric algorithm that can overcome these problems with more accurate
prediction and human opinion independency, without relying on the domain expert. The new algorithm is done by
merging together two techniques called computational approach and Winnow algorithm. This new algorithm also
uses more effective attributes than frequent words which is the most successful attribute in Stylometric prediction
this far. The effective attributes have been represented by the word pair and the trio. The experiments show that
the new algorithm produces superior prediction and even provides completely correct results.
The fact that pervasive systems are typically embedded and invisible, make them difficult for users to know
when these devices are collecting data. Consequently privacy appears as a major issue for pervasive computing
applications and several privacy models proposed for pervasive environments. The rapid growth of privacy
models in pervasive environments gives rise to the need for some standard benchmarks and evaluation models
to evaluate and compare these privacy models. In this paper, we review privacy evaluation models in pervasive
environments and present their evaluation results. Then we propose an evaluation model that evaluates privacy
models based on user control over private information, expressiveness of privacy policies, and unobtrusiveness of
privacy mechanisms as well as represents the model privacy level in a matrix. Finally we evaluate several privacy
models by using the proposed privacy evaluation model.