Abstract
We develop a framework for the detection of high-level events in a social network context, allowing us to identify abnormal or malicious behavior such as spamming. Additionally, we can classify users by analyzing their typical behavior while logged into a social network site. The processing of (real-time) events in our framework is done via an event detection language called ISEQL, which we adapt and extend to fit the requirements of a social network setting. We evaluate our framework experimentally, showing its effectiveness and efficiency.