automated bug classification using bayesian probabilistic approach
DESCRIPTION
TRANSCRIPT
RESEARCH POSTER PRESENTATION DESIGN © 2012
www.PosterPresentations.com
• Bug Classification - Precondition for Bug Fixation
• Miss-classification- Increases Fixation Time & Cost
• Correct Classification- Helps Software Maintenance
• ExTax Framework Provides Classified Samples• Bayesian Probabilistic Approach for
Classification• Comparison between Naïve Bayes & Bayes
Net
Motivation
• Subjective and Objective Interpretation of Bug Classes• Extensible Bug Classification, Seaman et al., ESEM,
2008• Manual Defect Analysis, Nakamura et al., ISESE, 2006• UNIX Security Bug Taxonomy, Aslam, Purdue University• Orthogonal Defect Classification, Chillarege et al., IEEE,
1992• ExTax Framework, Billah and Roy, SRLAB, UoS, 2012 • NBC on Bugzilla and Eclipse, Neelofar et al., CISIS, 2012
Background
• Common Vulnerabilities and Exposure (CVE)• MITRE Corporation• 61,894 samples, 25,350 classified samples• 5 groups, 22 classes, 64 attributes• Natural Language text for Bug Description.
Dataset
• Conditional Independence among Class Attributes• Naïve Bayes Classifier – Conditional Independence of
Attributes• Bayes Net – Conditional Dependence of Class
Attributes• Bayes Net Formulation for Bug Classes & Attributes• Conditional Probability Tables• Training, Model Selection, Classification
Hierarchical Bug Classification using Bayesian Probabilistic Approach
Evaluation Procedure
• Two Machine Learning Approaches – NBC and Bayes Net
• ExTax Classification Data for Evaluation• Leave-one-out Cross Validation Test• Performance Metrics – Precision, Recall, F-Score• NBC versus Bayes Net• Proposed Approach vs. Existing Approaches
Project Plan• Subject to strength and weakness of ExTax• Challenge to work with 22 classes, 64 attributes.
References[1] K. A. M. Billah and C. K. Roy. ExTax: A user driven
classification framework for extensible source code defect taxonomies. Technical report, University of Saskatchewan, Department of Computer Science, 2012.
[2] C. B. Seaman, F. Shull, M. Regardie, D. Elbert, R. L. Feldmann, Y. Guo, and S. Godfrey. Defect categorization: making use of a decade of widely varying historical data. In Proc. ESEM'08, ESEM '08, pages 149-157, New York, NY, USA, 2008. ACM.
LOGOMohammad Masudur Rahman, Shamima Yeasmin
Department of Computer Science, University of Saskatchewan, Saskatoon, SK, Canada
Hierarchical Bug Classification using Bayesian Probabilistic Approach
True classes
Retrieved classes
Fig. Bayes Net of NBC
Fig. Bayes Net of CVE group, class and attributes
Fig. Schematic Diagram of Proposed Approach
Table. CVE groups and classes
Table. Project Schedule
Fig. Precision & Recall