The following papers have been published for the Robust and Failure-Aware AI project<\/p>\n\n\n\n
Qi Zhang, Dian Chen, Lance M. Kaplan, Fellow, IEEE, Audun J\u00f8sang, Dong Hyun Jeong, Member, IEEE, Feng Chen, Member, IEEE, Jin-Hee Cho<\/p>\n\n\n\n
Abstract<\/strong>: Many studies have attempted to quantify uncertainty in model predictions to detector misclassification, but sometimes a single uncertainty value cannot fully capture misclassifications, especially when phishing or spam messages appear deceptively legitimate, and also lack interpretability of why the misclassification happens. This paper proposes X-MAP, an explainable framework that reveals topic-level patterns associated with incorrect predictions in text-based spam and phishing classification. We integrate SHapley Additive exPlanations (SHAP) with non-negative matrix factorization and distance-based analysis to derive topic profiles for reliably classified messages and measure how individual messages deviate from these profiles. Experiments on SMS spam and phishing email datasets show that misclassified messages exhibit 2 to 10 more times higher Jensen–Shannon divergence from the corresponding reliable groups than correctly classified ones, empirically validating our distance-based assumption. As a misclassification detector, X-MAP achieves up to 0.98 AUROC and reduces the false rejection rate at 9\\% true rejection to about 9% on positive predictions, while preserving the underlying classifier’s accuracy. When used as a repair layer on top of an Uncertainty Quantification based detector, X-MAP recovers a substantial portion of falsely rejected but correct predictions with minimal error leakage, providing interpretable, topic-level indicators of potential misclassification.<\/p>\n\n\n\n This paper is under review<\/p>\n","protected":false},"excerpt":{"rendered":" The following papers have been published for the Robust and Failure-Aware AI project X-MAP: eXplainable Misclassification Analysis and Profiling for Spam and Phishing Detection Qi Zhang, Dian Chen, Lance M. Kaplan, Fellow, IEEE, Audun J\u00f8sang, Dong Hyun Jeong, Member, IEEE, Feng Chen, Member, IEEE, Jin-Hee Cho Abstract: Many studies have attempted to quantify uncertainty in […]<\/p>\n","protected":false},"author":392,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"class_list":["post-54","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/wordpress.cs.vt.edu\/trustworthy\/wp-json\/wp\/v2\/pages\/54","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/wordpress.cs.vt.edu\/trustworthy\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/wordpress.cs.vt.edu\/trustworthy\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/wordpress.cs.vt.edu\/trustworthy\/wp-json\/wp\/v2\/users\/392"}],"replies":[{"embeddable":true,"href":"https:\/\/wordpress.cs.vt.edu\/trustworthy\/wp-json\/wp\/v2\/comments?post=54"}],"version-history":[{"count":5,"href":"https:\/\/wordpress.cs.vt.edu\/trustworthy\/wp-json\/wp\/v2\/pages\/54\/revisions"}],"predecessor-version":[{"id":93,"href":"https:\/\/wordpress.cs.vt.edu\/trustworthy\/wp-json\/wp\/v2\/pages\/54\/revisions\/93"}],"wp:attachment":[{"href":"https:\/\/wordpress.cs.vt.edu\/trustworthy\/wp-json\/wp\/v2\/media?parent=54"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}