Authors:
Ryu Watanabe
1
;
Takashi Matsunaka
1
;
Ayumu Kubota
1
and
Jumpei Urakawa
2
Affiliations:
1
KDDI Research, Inc., Saitama, Japan
;
2
KDDI Digital Security Inc., Tokyo, Japan
Keyword(s):
Security Measure, Vulnerability Management, Security Alert, Machine Learning.
Abstract:
The security alerts announced by various organizations can be used as an indicator of the severity and danger of vulnerabilities. The alerts are public notifications issued by security-related organizations or product/software vendors. The experts from such organizations determine whether it is a necessity of a security alert based on the published vulnerability information, threats, and publicized damages caused by the attacks to warn the public of high-risk vulnerabilities or cyberattacks. However, it may take some time between the disclosure of the vulnerability and the release of a security alert. If this delay can be shortened, it will be possible to guess the severity of the vulnerability earlier. For this purpose, the authors have proposed a machine learning method to predict whether a disclosed vulnerability is severe enough to publicize a security alert. In this paper, our proposed scheme and the evaluation we conduct to verify its accuracy are denoted.