This paper presents a study of existing threat detection techniques in cloud computing, together with an experimental evaluation of a subset of them. We consider the threats defined in the Cloud Security Alliance (CSA) report as well as the techniques for their detection, starting from classical signature-based approaches and finishing with recent machine learning based techniques. This paper also contains an analysis of original results presented in international conferences, published as journal papers, Internet resources, and standards. The main contributions of the study include: (1) providing a closer relationship between top threats in cloud computing and known detection techniques; (2) evaluating existing detection techniques concerning cloud computing principles and security challenges nowadays; and (3) reviewing commonly utilised datasets and their association with threats in the last five years. As existing detection techniques tend to target specific threats (or their groups), we also present the experimental evaluation of the applicability of known detection approaches against non-targeted threat groups