For the past few years, malware or also known as malicious code is seen as one of the biggest threats of the cyber attacks. It has caused lot of damages, loss of money and productivity to many organizations and end users. Malicious code can be divided into many categories such as viruses, worms and trojan horses. Each of these categories has it owns implications and threats, and trojan horse has been chosen as the domain of this research paper. Prior to the formation of a new trojan horse detection model, an in-depth study and investigation of the existing trojan horse classification is presented in this paper. Surprisingly, not much research related with trojan horse has been done. On 16th January 2013, Troj/Invo- Zip has caused chaos by masquerading as an invoice from Europcar and spreading via email. Therefore, in this research paper, a new trojan horse classification called Efficient Trojan Horse Classification (ETC) is developed. This ETC later is used as a basis to build a model to detect trojan horse efficiently. The methods used to develop the ETC are the static and dynamic analyses. As for the dynamic analysis, cuckoo sandbox has been integrated to speed up the analysis and reverse engineering processes.