U.S. Science and Technology Directorate seeks ideas for secure generation of synthetic data

The Department of Homeland Security’s (DHS) Science and Technology Directorate (S&T) this week solicited the public for ideas on synthetic data generation. 

This data would model and replicate the shape and patterns of real data, but in a more secure fashion the department could use to train machine learning models. The advantage here is that synthetic data would have to worry about inadvertently utilizing sensitive information, like personal details. Ideally, it would have a variety of applications and provide help in ways such as:

1. Support structured and unstructured data types
2. Artificially generate synthetic data using techniques that seek to understand the shape and patterns of real data
3. Replicate data sets with desired statistical attributes
4. Remove and/or mitigate bias in synthetic data
5. Generate data of sufficient quality to train machine learning (ML) models
6. Verify data fidelity
7. Discern how different generation techniques may be suited to different types of data
8. Prevent real data being reverse engineered from synthetic data
9. Support structured and unstructured data types
10. Artificially generate synthetic data using techniques that seek to understand the shape and patterns of real data

“It is crucial for DHS to effectively navigate today’s complex privacy landscape and employ innovative ideas and next generation technology techniques to do so,” Melissa Oh, Managing Director of S&T’s Silicon Valley Innovation Program (SVIP), said. “The ability to generate synthetic data at scale is necessary to protect and preserve data privacy, as well as safeguard civil rights and liberties.”

In return, participants would be eligible for up to $1.7 million of non-dilutive funding stretched over four phases of development and adaptation of commercial technologies for homeland security use cases. Companies will have until April 10, 2024 to apply.