Fortinet, Inc.

Arcsoft Corporation Limited

Juneng Zheng

Guoyi Yan

Hongbin Lu

Systems and methods for preprocessing data to facilitate DLP pattern matching are provided. An input string is received by a Data Leak Prevention (DLP) system. The input string is converted by the DLP system into a fixed string pattern. The conversion is performed based on one or more of multiple class definitions, including a digit class, a letter class and a symbol class. A determination is then made by the DLP system regarding whether the input string contains potential sensitive data to which a full-match is to be applied by matching the fixed string pattern against one or more stored target strings representative of sensitive data.

Pattern matching for data leak prevention

Systems and methods for high performance IDS/IPS with efficient metadata filtering are provided. According to one embodiment, a signature database of an IDS/IPS is configured with multiple metadata signatures. A pre-match engine identifies a candidate packet of network traffic received by the IDS/IPS for full-feature match processing by: (i) categorizing the metadata signatures based on characteristics thereof; and (ii) processing and filtering a first set of the metadata signatures that forms part of a hash key based metadata signature category. The hash key based metadata signature category represents a category resulting from the categorization and each of the first set of metadata signatures is associated with a fixed unique hash key based on which respective metadata signatures are matched with the received network traffic to identify the candidate packet. Finally, a potential intrusion is identified by performing the full-feature match processing on the candidate packet based on full-feature match signatures.

Filtering of metadata signatures

Methods and systems for determining consecutive matches are provided. According to one embodiment, a class definition and a data stream are received by a network security device. The data stream is partitioned into multiple data blocks each containing N data segments. Each data block is processed in parallel to compute: (i) a value (F) indicating whether every data segment value meets the class definition; (ii) a value (L) indicating a number of consecutive data segment values meeting the class definition starting from the left; (iii) a value (M) indicating a maximum number of consecutive data segment values meeting the class definition; and (iv) a value (R) indicating a number of consecutive data segment values meeting the class definition starting from the right. Corresponding values for each data block are then aggregated to determine a maximum number of consecutive data segment values meeting the class definition for the entire data stream.

Calculating consecutive matches using parallel computing

Systems and methods for filtering unsafe content by a network security device are provided. According to one embodiment, a network security device captures network traffic and extracts a media file from the network traffic. The network security device then determines the presence of a hidden data item embedded in the media file in a machine-readable form. When such a hidden data item is identified, the network security device performs one or more actions on the media file based on a predefined security policy.

Filtering hidden data embedded in media files

Systems and methods for preprocessing data to facilitate DLP pattern matching are provided. An input string is received by a Data Leak Prevention (DLP) system. The input string is converted by the DLP system into a fixed string pattern. The conversion is performed based on multiple class definitions, including a digit class, a letter class and a symbol class. A determination is then made by the DLP system regarding whether the input string contains potential sensitive data to which a full-match is to be applied by matching the fixed string pattern against one or more stored target strings representative of sensitive data.

Systems and methods for rating of signature patterns are provided. According to one embodiment, a frequency of occurrence is determined by a network security system of each of multiple patterns within a pattern database containing a set of candidate patterns from which a set of patterns or sub-patterns thereof will be selected for inclusion within a pre-match list. For each pattern, the network security device determines whether a length of the pattern exceeds a pre-defined length; and, if so, cuts the pattern to generate multiple sub-patterns having the pre-defined length. A rating for each pattern or, as the case may be, each sub-pattern is then determined by the network device based on any or a combination of the frequency of occurrence of the pattern within the pattern database, the length of the pattern or the sub-pattern and a measure of redundancy within the pattern or sub-pattern.

Rating of signature patterns for pattern matching

Systems and methods for filtering unsafe content at a network security appliance are provided. According to one embodiment, a network security appliance captures network traffic and extracts a media file from the network traffic. The network security appliance then determines the presence of a hidden data item embedded in the media file in a machine-readable form. When such a hidden data item is identified, the network security appliance performs one or more actions on the media file based on a predefined security policy.

A method generates a composite image by painting unfilled regions between micro-objects with other micro-objects to create a multiple layer effect. Another method generates a composite image from as few as one micro-object by changing the average color and/or intensity of the micro-object to match that of an area on the source image. Yet another method quickly generates a preview image of a composite image by saving the micro-objects in multiple resolutions so that the micro-objects at the small resolution can be quickly scaled to build a preview image of the composite image.

Growing community of inventors

Fremont, CA, United States of America

Juneng Zheng