A method of determining culture related users using computation of correlation

UNCG Author/Contributor (non-UNCG co-authors, if there are any, appear on document)
Rui Da (Creator)
Institution
The University of North Carolina at Greensboro (UNCG )
Web Site: http://library.uncg.edu/
Advisor
Jing Deng

Abstract: The provision of security on most of computer networks is based on the obtaining and exchange of a unique key between the communicating parties. It is, however, difficult to come up with a truly unique and random secret between two parties with the help from physical randomness. In this work, we focus on the problem of unique random number generation or derivation between users in online social networks. As a result of rapid development of Internet, online social networks provide a vast set of different user comments on different products and services. Such comments can inherently reflect the mindset or cultural background of those people who wrote them and it is possible to derive some unique randomness from such texts with some maneuver. We select movie reviews as the sandbox for our investigation. To manage textual content and search for certain hidden relations, the methodologies of text matching are studied. By looking the similarities of movie reviews from different users, we can refer insights into the cultural background and even predict future preferences from past comments. We present all of our findings here to aspire further investigation. We have investigated the correlation of movie reviews and studied the values of different weight assignments to the sentence and word relation. According to our results, synonym relations are the dominant positive association that impacts correlation value. We calculate correlation between review sets containing multiple reviews to avoid randomness. These correlations have then been used to evaluate and derive a unique random number. We target at a single review, and put it together with other reviews to obtain correlation values from different pairs of reviewers. Then the correlation value is binning to a 1-bit binary number. Through such a simplified extraction, a unique random number can be generated by repeating the process of binning. Such unique random number is able to facilitate to secure information exchanges between the users. In our future work, we will explore such correlations to generate a practically usable unique secret for secret keys.

Additional Information

Publication
Thesis
Language: English
Date: 2015
Keywords
Key generation, Movie review correlation, Text matching
Subjects
Computer networks $x Security measures
Random number generators $x Computer programs

Email this document to