Statistical Learning Methods for Information Security: Fundamentals and Case Studies

Abstract

One of the most traditional methods for information security can be as easy as sequence matching, such as the signature-based methods for virus detection. However, it is now well-accepted that the signature-based methods are no longer satisfactory solutions for many security problems. The signature is usually too rigid, resulting in detection that is hard to adjust and easy to bypass. Statistical learning approaches can complete the puzzle to form an integrated defense system. Numerous statistical learning methods have been proposed in the last couple of decades for various applications. To solve information security problems statistically, we need to carefully choose appropriate statistical learning methods and evaluation procedures so that what seems to be a meaningful and effective method in terms of the statistical analysis can also be beneficial when the method is deployed to the real world. This paper aims to give an introductory and as self-contained as possible overview for how to correctly and effectively apply statistical methods to information security problems. We also demonstrate a couple of applications of the statistical learning methods on the problems of botnet detection and account security.

Citation

Hsing-Kuo Pao, Yuh-Jye Lee, and Chun-Ying Huang, "Statistical Learning Methods for Information Security: Fundamentals and Case Studies," Applied Stochastic Models in Business and Industry, Vol. 31, Issue 2, pp. 97--113, March 2015.

Bibtex

@article{pao15:asmbi, author = {Hsing-Kuo Pao and Yuh-Jye Lee and Chun-Ying Huang}, title = {Statistical Learning Methods for Information Security: Fundamentals and Case Studies}, journal = {Applied Stochastic Models in Business and Industry}, volume = {31}, number = {2}, pages = {97--113}, month = {March}, year = {2015} }

Download