USSOCOM seeks disinformation campaign detection tech
The U.S. Special Operations Command posted a request for information for a Misinformation, Disinformation and Mal-Information Campaign Detection Prototype. Responses are due by 12:00 p.m. Eastern on January 13.
The Headquarters, United States Special Operations Command (HQ USSOCOM) is conducting market research in accordance with FAR 15.201. THIS IS NOT A REQUEST FOR PROPOSAL (RFP).
United States Special Operations Command (USSOCOM) has a requirement for a Commercial Off the Shelf (COTS) software solution, to be used as a prototype for use of understanding the information environment that can detect misinformation, disinformation and mal-information campaigns in near to real-time to directly support information operations within Special Operations Command. The resulting analysis will be surfaced on a user interface and contextualized by the relevant narratives and network of accounts. This software must leverage an intuitive cloud-hosted user interface software which provides processing and analyzing multi-modal social media and web data and has the programmatic ability to dissect and categorize information sectors. The government requires a prototype data pipeline that will identify viral and trending content for threat assessments and score data with a ranking system that highlights the likelihood of being fake or deceptive and display the information on the user interface.
The prototype will use a combination of deep learning, natural language processing, and dynamic network analysis to detect and examine the cross-platform spread of disinformation (in the form of images, video, and text) over vast and disparate data sources and inform teams of information which is intended to be deceptive. The vendor must address anomaly detection, deep fake, foreign influence and anticipatory analytics, leveraging deep learning, natural language processing, and proven machine learning development and capabilities.
The resulting analysis will be surfaced on a user interface and contextualized by the relevant narratives and network of accounts. The system is designed to improve over time by incorporating user feedback and adapt to the rapidly changing tactics of an adversary.
The vendor must use network graph analysis to narrow focus to a relatively small target subset, which will result in a reduced computational load for the deep fake detectors. The analysis will be required to incorporate analyst feedback for a human-in-loop pattern to improve the models over time, leveraging supervised machine learning. In addition, because the analysis will include not only the original content but also the fuller network context, the developed prototype is expected to deliver a much richer and nuanced set of results to better understand and identify misinformation/disinformation across the information environment.
Full information is available here.