The challenging task in cyber space is to detect malicious URLs. The websites pointed by the malicious URLs injects malicious code into the client machine or steals the crucial information. As detecting a phishing URL is a challenging task, it is essential to enhance detection techniques against the emerging attacks. The most of the existing approaches are feature based and cannot detect dynamic attacks. Mostly the attacker uses the input form, active content and embeds at the rate symbol in URL for malicious attack. To detect this attack, a Behaviour based Malicious URL Finder (BMUF) algorithm is proposed. It analyzes the behaviour of the URL. The FSM based state transition diagram is used to model the URL behaviour into various states. The state transition from initial to final state is used for classification. This approach tests the genuine and malicious behavior of the URL based on the responses to the user. It accurately detects the nature of the URL