Facebook spammers often use Facebook groups to propagate spam. A Facebook group member can invite his friends to join the group without the invitees’ permission. Such a convenient invitation mechanism allows a spammer to add compromised user accounts and their friends to a Facebook group created by the spammer. Then, whenever a new message is posted on the group’s wall, every member will receive a notification of the post automatically. This automatic mechanism applies to all group members no matter whether they know or have ever visited this group. As a result, a spammer can easily create a Facebook group to spread spam. A Facebook group which is created to scatter spam is called a spamming group. Even though detection of e-mail spam or web-based spam has been developed for a long period of time, current Facebook mechanisms still cannot efficiently remove spamming groups. Only 14 of 346 spamming groups we monitored were deleted by Facebook in April 2014. Most of the above 346 spamming groups exist for at least five months during our experimental time. Therefore, it becomes an important issue to develop a new solution to identify spamming groups. In this paper, we propose a behavior-based spamming group detection approach for Facebook, called Itus. Itus has an auxiliary crawling Chrome extension to collect and extract features from Facebook groups. These features include relationships between members and their relevant social activities. These features are used for training Itus support vector machine, a machine learning based classifier that can identify a spamming group efficiently. Experimental results shows that the best total detection error rate of Itus is 3.27%.