DocMark: A Rule-based and Probabilistic Document Marking System

ABSTRACT: In order to prevent loss of sensitive data, organizations need to be able to tag data with classification and releasability metadata. Currently classification is a manual and labor-intensive process, which slows the dissemination of sharable data. ATC-NY will develop DocMark, a product to automatically determine classification and dissemination metadata and record it in standard formats for review by a human subject matter expert. DocMark will perform both file marking and portion marking (e.g., frames in a video stream). It will provide graphical front-ends for configuration and review, and APIs for marking individual data objects or collections of them. By automatically performing many of the steps that human beings now perform manually, DocMark will dramatically improve the speed at which data can be marked. BENEFIT: Many legacy military and intelligence sensors produce data objects that are not marked with the classification and releasability metadata needed for secure sharing. Marking is required both for newly created data objects and for repositories of unmarked objects. DocMark will convert marking from a manual process to a semi-automated one, in which the job of the human being is to review and occasionally correct a marking provided by DocMark. The first target markets of DocMark are the intelligence and military communities, who have an urgent need to disseminate data to the personnel who need it. The business community is now discovering its own need for classification, which helps prevent data loss by clarifying the difference between highly sensitive objects and those that can be freely disseminated. DocMark will be available as a commercial product; ATC-NY will also provide training, support, and integration/customization services.