Neuro-Symbolic Integration for Detecting Phishing Attacks
Abstract
A zero-day phishing attack exploits a potentially serious security vulnerability that people may be unaware of. Considering the fatality of phishing attacks that are emphasized by many organizations, the inductive learning approach using reported malicious URLs has been verified in the field of deep learning. However, the deep learning-based method, mainly focused on fitting a classification task via historical URL observation, has a limitation of recall due to the characteristics of zero-day attack. To model the nature of a zero-day phishing attack in which URL addresses are generated and discarded immediately, an approach that utilizes expert knowledge is promising. We introduce the integration method of deep learning and logic programmed domain knowledge to inject the real-world constraints. The research group designs neural and logic classifiers and develop the joint learning method of each component based on the neuro-symbolic integration.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 10, 2022
- Source ID
- FA23862114085XX49
Entities
People
- Sung-Bae Cho
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- Yonsei University