Nowadays, spam has become serious issue for computer security, because it becomes a main source for disseminating threats, including viruses, worms and phishing attacks. Currently, a large volume of received emails are spam. Different approaches to combating these unwanted messages, including challenge response model, whitelisting, blacklisting, email signatures and different machine learning methods, are in place to deal with this issue. These solutions are available for end users but due to dynamic nature of Web, there is no 100% secure systems around the world which can handle this problem. In most of the cases spam detectors use machine learning techniques to filter web traffic. This work focuses on systematically analyzing the strength and weakness of current technologies for spam detection and taxonomy of known approaches is introduced.