X-means: Extending K-means with Efficient Estimation of the Number of Clusters
ICML '00 Proceedings of the Seventeenth International Conference on Machine Learning
Detecting and analyzing automated activity on twitter
PAM'11 Proceedings of the 12th international conference on Passive and active measurement
| Hi-index | 0.00 |
There are many anti-spam techniques available today. However, spammers evolve mass mailing techniques in order to circumvent these countermeasures. One example of such evolutionally advanced spammers is observed in email services offered by Japanese mobile phone service providers. Because they have been enforcing very strict anti-spam filters, commonly used mass mailing techniques such as spam botnets are becoming less effective, and spammers thus have to evolve their technologies. In order to understand such evolutionally advanced spam-sending hosts' behaviors, we collected and analyzed their traffic flow data retrieved at a backbone network in the real commercial network of one of the largest mobile phone service providers in Japan, which has over 30 million customers. In this paper, we first show that many of the existing anti-spam techniques are not effective against advanced spammers, and then reveal that such advanced spammers have distinctive time-series behavioral characteristics that have the potential to be exploited in developing new mitigation techniques and predicting their behavior in the future.