Threat categorization on CVE descriptions using text classification
View/
Keywords
Threats ; CVE ; CWE ; CAPEC ; Natural language processing ; Text classificationAbstract
The goal of this thesis is to classify CVEs into potential threats based on the descriptions provided using text classification, to then be used in a vulnerability-based risk analysis for Cyber-Physical Systems and more specifically Intelligent Transportation Systems. Because not all CVEs provide related CWEs, or their CWE entries are too generic, such as NVD-CWE-Other, a mapping of threats to CWEs and CAPECs, as used by the CitySCAPE Project's Risk Modelling Tool, is not always possible. This thesis proposes using text classification in order to identify those threats based on CVEs that can be mapped to CWEs and CAPECs. An accuracy of over 90% was achieved for each of the 16 threats across 111,384 CVE entries computed on a 10-fold cross validation.