The increase in phishing attacks through email and short message service (SMS) has shown no signs of deceleration. The first thing we need to do to combat the ever-increasing number of phishing attacks is to collect and characterize more phishing cases that reach end users. Without understanding these characteristics, anti-phishing countermeasures cannot evolve. In this study, we propose an approach using Twitter as a new observation point to immediately collect and characterize phishing cases via e-mail and SMS that evade countermeasures and reach users. Specifically, we propose CrowdCanary, a system capable of structurally and accurately extracting phishing information (e.g., URLs and domains) from tweets about phishing by users who have actually discovered or encountered it. In our three months of live operation, CrowdCanary identified 35,432 phishing URLs out of 38,935 phishing reports. We confirmed that 31,960 (90.2%) of these phishing URLs were later detected by the anti-virus engine, demonstrating that CrowdCanary is superior to existing systems in both accuracy and volume of threat extraction. We also analyzed users who shared phishing threats by utilizing the extracted phishing URLs and categorized them into two distinct groups - namely, experts and non-experts. As a result, we found that CrowdCanary could collect information that is specifically included in non-expert reports, such as information shared only by the company brand name in the tweet, information about phishing attacks that we find only in the image of the tweet, and information about the landing page before the redirect. Furthermore, we conducted a detailed analysis of the collected information on phishing sites and discovered that certain biases exist in the domain names and hosting servers of phishing sites, revealing new characteristics useful for unknown phishing site detection.

In this paper, we propose a method to reduce the labeling cost while acquiring training data for a malicious domain name detection system using supervised machine learning. In the conventional systems, to train a classifier with high classification accuracy, large quantities of benign and malicious domain names need to be prepared as training data. In general, malicious domain names are observed less frequently than benign domain names. Therefore, it is difficult to acquire a large number of malicious domain names without a dedicated labeling method. We propose a method based on active learning that labels data around the decision boundary of classification, i.e., in the gray area, and we show that the classification accuracy can be improved by using approximately 1% of the training data used by the conventional systems. Another disadvantage of the conventional system is that if the classifier is trained with a small amount of training data, its generalization ability cannot be guaranteed. We propose a method based on ensemble learning that integrates multiple classifiers, and we show that the classification accuracy can be stabilized and improved. The combination of the two methods proposed here allows us to develop a new system for malicious domain name detection with high classification accuracy and generalization ability by labeling a small amount of training data.