US Department of Homeland Security Solicits Solutions for Synthetic AI/ML Training Data.

The U.S. Department of Homeland Security (DHS) Science and Technology Directorate announced today that it is soliciting ideas from the public for solutions to generate synthetic data for training artificial intelligence and machine learning models. Accepted submissions are eligible to receive up to US$1.7M in funding to develop and adapt commercial technologies for DHS’s use. 

The request recognizes the potential privacy and cybersecurity risks associated with using “live” data for training these models, while also recognizing the significant challenges of creating synthetic data that accurately models real-world data without bias (although the challenges of bias in live data are still significant). DHS explicitly stated that one capability the proposed solutions should provide includes “removing and/or mitigating bias in synthetic data” and recognizes that the solutions should also include privacy preserving capabilities that meet the mission needs of DHS and its various agencies, such as the Cybersecurity & Infrastructure Security Agency (CISA) and the DHS Privacy Office. 

Applications are due by April 10, 2024 and there will be a hybrid “industry day” in Durham, North Carolina to discuss the solicitation and answer questions. 

The Department of Homeland Security (DHS) Science and Technology Directorate (S&T) announced a new solicitation seeking solutions to generate synthetic data that models and replicates the shape and patterns of real data, while safeguarding privacy and mitigating security harms. Synthetic data is important for DHS because it allows the Department to train machine learning models using synthetic data when real-world data is not available, or when using it would pose privacy and security risks, particularly if the real-world data includes sensitive information, such as personally identifiable information (PII).

View referenced article