A team from the U.S. Intelligence Advanced Projects Agency (IARPA) — which is the research arm of the U.S. director of national intelligence — is seeking out technologies that will authenticate the authorship of both human and machine generated documents.

The aim of the Human Interpretable Attribution of Text using Underlying Structure (HIATUS) program is to create explainable linguistic fingerprints that can be attributed to specific authors, thereby protecting an author’s privacy.

According to the researchers, authorship privacy, or authorship obfuscation, references methods of modifying text so as to eliminate features that are unique to a specific author. Such technologies designed to identify the hallmarks of an author's so-called linguistic fingerprint could potentially be used to obscure such fingerprints, shielding individuals and groups whose writing — if linked to those groups or persons — could put them in danger.

The researchers suggest that such technology could also pave the way for improving the understanding of malicious online information campaigns, fighting human trafficking and identifying counterintelligence risks.

In addition to creating technologies that automatically identify and produce verifiable linguistic fingerprints of both human and machine authors, the HIATUS team is planning to develop an encoder that can compare texts from the same author.

Submissions to the HIATUS program were due on April 18, 2022.

To contact the author of this article, email mdonlon@globalspec.com