Abstract: Features of sample files that are known to be normal are extracted by random projection. The random projection values of the sample files are used as training data to generate one or more anomaly detection models. Features of a target file being inspected are extracted by generating a random projection value of the target file. The random projection value of the target file is input to an anomaly detection model to determine whether or not the target file has features that are novel relative to the sample files. The target file is declared to be an outlier when an anomaly detection model generates an inference that indicates that the target file has novel features.

Abstract: A locality-sensitive hash value is calculated for a suspect file in an endpoint computer. A similarity score is calculated for the suspect hash value by comparing it to similarly-calculated hash values in a cluster of known benign files. A suspiciousness score is calculated for the suspect hash value based upon similar matches in a cluster of benign files and a cluster of known malicious files. These similarity score and the suspiciousness score or combined in order to determine if the suspect file is malicious or not. Feature extraction and a set of features for the suspect file may be used instead of the hash value; the classes would contain sets of features rather than hash values. The clusters may reside in a cloud service database. The suspiciousness score is a modified Tarantula technique. Matching of locality-sensitive hashes may be performed by traversing tree structures of hash values.

Abstract: A system for evaluating a target file includes an endpoint computer that receives similarity digests of legitimate files, receives a target file, and generates a similarity digest of the target file. The endpoint computer determines whether or not the target file is legitimate based on a comparison of the similarity digest of the target file against the similarity digests of the legitimate files. The system further includes a backend computer system that receives the legitimate files, generates the similarity digests of the legitimate files, and provides the similarity digests of the legitimate files to the endpoint computer.