Proposal / Submission Type
Peer Reviewed Paper
Location
Richmond, Virginia
Start Date
31-5-2012 10:30 AM
Abstract
A multi-parameter sensitivity analysis of a Bayesian network (BN) used in the digital forensic investigation of the Yahoo! email case has been performed using the principle of ‘steepest gradient’ in the parameter space of the conditional probabilities. This procedure delivers a more reliable result for the dependence of the posterior probability of the BN on the values used to populate the conditional probability tables (CPTs) of the BN. As such, this work extends our previous studies using singleparameter sensitivity analyses of BNs, with the overall aim of more deeply understanding the indicative use of BNs within the digital forensic and prosecutorial processes. In particular, we find that while our previous conclusions regarding the Yahoo! email case are generally validated by the results of the multi-parameter sensitivity analysis, the utility of performing the latter analysis as a means of verifying the structure and form adopted for the Bayesian network should not be underestimated.
Keywords: Bayesian network; digital forensics; multi-parameter sensitivity analysis; steepest gradient.
Scholarly Commons Citation
Overill, Richard E.; Zhang, Echo P.; and Chow, Kam-Pui, "Multi-Parameter Sensitivity Analysis of a Bayesian Network from a Digital Forensic Investigation" (2012). Annual ADFSL Conference on Digital Forensics, Security and Law. 10.
https://commons.erau.edu/adfsl/2012/thursday/10
Included in
Computer Engineering Commons, Computer Law Commons, Electrical and Computer Engineering Commons, Forensic Science and Technology Commons, Information Security Commons
Multi-Parameter Sensitivity Analysis of a Bayesian Network from a Digital Forensic Investigation
Richmond, Virginia
A multi-parameter sensitivity analysis of a Bayesian network (BN) used in the digital forensic investigation of the Yahoo! email case has been performed using the principle of ‘steepest gradient’ in the parameter space of the conditional probabilities. This procedure delivers a more reliable result for the dependence of the posterior probability of the BN on the values used to populate the conditional probability tables (CPTs) of the BN. As such, this work extends our previous studies using singleparameter sensitivity analyses of BNs, with the overall aim of more deeply understanding the indicative use of BNs within the digital forensic and prosecutorial processes. In particular, we find that while our previous conclusions regarding the Yahoo! email case are generally validated by the results of the multi-parameter sensitivity analysis, the utility of performing the latter analysis as a means of verifying the structure and form adopted for the Bayesian network should not be underestimated.
Keywords: Bayesian network; digital forensics; multi-parameter sensitivity analysis; steepest gradient.
