Methods and devices are described for detecting code clones in a software program. The source code of the software program is processed into a group of n-gram representations. A clone index is generated for each respective code portion defined in the normalized source code, where each clone index includes a feature vector encoding features of the respective code portion based on the n-gram representations corresponding to the respective code portion. Comparison of the clone indexes is used to detect a code clone, based on matching the feature vectors of the clone indexes.