This disclosure relates to detection of viruses. In one implementation, a virus detection method performed by a terminal device is disclosed. The method may include obtaining a full characteristic features of a first file, the first file being a sample file collected by a client application of the terminal device, and the full characteristic features of the first file being used to uniquely determine all components of the first file, obtaining a first binary hash sequence based on the full characteristic features of the first file and by using a preset hash algorithm, calculating similarity parameters between the first binary hash sequence and reference binary hash sequences in an antivirus database, each binary hash sequence in the antivirus database corresponding to one type of virus-infected file, and determining the first file as a virus-infected file when at least one of the similarity parameters is less than a preset similarity threshold.