DHS S&T awards contracts to further privacy enhancing synthetic data

The US Department of Homeland Security (DHS) Science and Technology Directorate (S&T) announced that it has awarded contracts under its Silicon Valley Innovation Program (SVIP) to four start-up businesses to work on developing privacy enhancing synthetic data generation capabilities.

The capabilities are meant to model and replicate the shape and patterns of data while safeguarding privacy and mitigating security risk.

SVIP utilizes the Innovation Other Transaction Solicitation – SVIP OTS 70RSAT21R00000006 – to engage with organizations that do not usually work with the government to carry out prototype projects and possibly transition successful projects to production.

The four companies that were awarded the contracts are Singapore-based company Betterdata, Boston, Massachusetts-based DataCebo, Vienna, Austria-based MOSTLY AI, and San Ramon, California-based Rockfish Data.

“Awarding these contracts are vital because startups are uniquely positioned to offer agile, creative approaches that can help the department address complex challenges like data privacy and security in groundbreaking ways,” said Melissa Oh, S&T’s SVIP Managing Director. “Sharing sensitive data that can be used for analytics, testing, and training machine learning models across organizational boundaries is highly challenging,”

DHS said the four contracts were awarded in response to its earlier announcement about the Synthetic Data Generator solicitation, which was issued in partnership with DHS’s Cybersecurity & Infrastructure Security Agency (CISA) and the DHS Privacy Office.

The solicitation was issued “to address a critical need for DHS capabilities that support collaborative use of data and mitigate privacy and security risks inherent in the sharing of sensitive data,” DHS explained, emphasizing that “synthetic data is crucial for DHS because it enables the department to train machine learning models in situations where real-world data is either unavailable or poses privacy and security concerns.”

“CISA is eager to work with SVIP and the startup ecosystem as we strive to support the identification, development, and use of privacy-enhancing technologies,” said CISA Associate Chief of Strategic Technology Dr. Garfield Jones. “Our investment in these technologies and collaboration with industry partners enhances our own agency operations, allows us to support the overall privacy ecosystem and its stakeholders, as these technologies become a part of daily life.”

DHS said that each awardee was “selected through a highly competitive process,” and may be eligible for up to $1.7 million across four SVIP phases.

Under the first phase DHS said the four companies “presented innovative solutions that have the potential to provide immediate impact” to the department.

The four phases are: proof of concept; demonstration of a pilot-ready prototype; functional and red team testing; and pilot/testing in an operation environment.

S&T awarded $196,260 to Betterdata, which is leveraging its Large Synthetic Model capabilities to generate synthetic data that is statistically accurate when real data is unavailable, or low in volume, while enhancing data auditing using differential privacy. The solution supports tabular, sequential, relational and text data modalities.

DataCebo was given awarded $196,920 to continue development of AI-generated synthetic data using its Synthetic Data Vault E platform, which allows developers to build, deploy, and manage sophisticated generative AI models for enterprise-grade applications when real data is limited or unavailable.

MOSTLY AI was awarded $196,800 to develop its generative AI Platform to create accurate and private tabular synthetic data to support model training, analytics, and testing use cases. DHS said the company’s “platform has unique capabilities around generating fair synthetic data to combat bias in synthetic data generation.”

Finally, S&T awarded $199,300 to Rockfish Data, which has developed a high fidelity and privacy-preserving generative data platform that automatically adapts to diverse operational datasets and enables flexible generation for myriad use cases. Based on foundational research performed by Carnegie Mellon University, the Rockfish platform helps customers to accelerate their AI transformation and improve security, resilience, and innovation in operations.

Article Topics

biometrics  |  data privacy  |  DHS S&T  |  government purchasing  |  synthetic data  |  U.S. Government