FY26-0251 JOFOC Cignal Synthetic Data Generation.pdf

JUSTIFICATION AND APPROVAL FOR OTHER THAN FULL AND OPEN COMPETITION

(41 U.S.C. 3304(a)(1))

FY26-0251

Pursuant to the requirements of the Competition in Contracting Act (CICA) as implemented by the Federal Acquisition Regulation (FAR) 6.103 (FAR Class Deviation 25-111), and in accordance with the requirements of RFO 6.104, the justification for the use of the statutory authority under RFO 6.103 is justified by the following facts and rationale required under RFO 6.104-1 as follows:

1. Agency and Contracting Activity. The Department of Homeland Security (DHS) Science & Technology Directorate (S&T) proposes to enter into a contract on a basis other than full and open competition.

2. Nature and/or Description of the Action being Approved.

(a) Nature of action

DHS S&T intends to procure on a sole source basis maturation of an application to develop synthetic images for development of detection algorithms on checkpoint security screening systems.

(b) Name and address of the contractor.

Cignal LLC, 7991 E Back Mountain Road, Reedsville, Pennsylvania 17084-8823

(c) Contract type Firm Fixed Price with payable milestones

Note: If the products or services under the justification were previously awarded under another contract, also provide the period of performance end date for that contract(s).

1 In this document, citations to RFO Parts 6, 10, 15, 25 are citations specifically to FAR Class Deviation 25-11 for FAR Part 6, FAR Class Deviation 25-06 for FAR Part 10, FAR Class Deviation 26-08 R1 for FAR Part 15, and FAR Class Deviation 25-09 R1 for FAR Part 25, which were issued as part of the Revolutionary FAR Overhaul (RFO), implemented in response to Executive Order 14275, Restoring Common Sense to Federal Procurement, signed April 15, 2025.

Contract PoP Value Type of Funds Year of Funds Solicitation Number

70RSAT20T0000019 05/04/2020 - 02/04/2021 $142,465.20 RD&I FY19 70RSAT18R00000024

70RSAT21T00000010 07/28/2021 - 04/27/2022 $199,840.00 RD&I FY19 SVIP Follow-on Phase 2

70RSAT22T00000014 09/26/2022 - 04/09/2024 $499,959.72 RD&I FY22 SVIP Follow-on Phase 3

70RSAT24T00000023 09/3/2024 - 06/02/2026 $499,965.60 RD&I FY23 + FY24 SVIP Follow-on Phase 4

(d) Type of funding: Research & Development

(e) Year of funding. FY26

(f) Solicitation number. TBD

(g) Background information about the requirement.

DHS S&T issued solicitation 70RSAT18R00000024 in 3QFY19 under the title “Object Recognition and Adaptive Algorithms in Passenger Property Screening.” The contractor proposed developing a novel capability to simulate raw X-ray image data and millimeter wave image data that screening systems initially produce, and this image data is used to develop detection algorithms that are a required element of checkpoint screening systems acquired by the Transportation Security Administration (TSA). Screening systems deployed at checkpoints use an Automated Target Recognition (ATR) algorithm, and these algorithms must be developed based on image data from the Computed Tomography (CT)/X-ray or millimeter wave based systems. ATR algorithm developers must invest considerable resources to build hundreds of passenger carry-on bags and recruit mock passengers to be screened by the systems to generate the image data on which an ATR algorithm can be built. Upon advancing such algorithms to an acceptable point, Government laboratories take over the role of providing bags and equipping mock passengers with realistic, and hazardous, threats, for the final stages of algorithm development, which no original equipment manufacturer can do, as they do not have access to threat materials. DHS S&T required a capability to produce synthetic image data which has all the signal and data attributes from real screening scans, potentially saving significant time and resources by both original equipment manufacturers’ algorithm developers as well as by Government laboratories.

The contractor Cignal was awarded an initial “Phase I” and subsequently demonstrated satisfactory development of novel their synthetic image generation in follow on phases through June 2026. This requirement is to fund the final stages of maturation of the synthetic image data generation tool and support plans for transition to TSA, other DHS Components, and commercialization for private sector algorithm developers.

3. Description of Supplies/Services.

In this research and development effort, the Contractor shall continue the development of and demonstrate software tools that support synthetic data generation suitable for public algorithm development (non-sensitive) and Government development, testing, and evaluation (sensitive). It is anticipated that the contractor shall collaborate with potential end users at S&T, TSA, and elsewhere to validate the tools and generated images for suitability for machine learning algorithm development, training, testing, and evaluation over the course of the contract.

4. Identification of Statutory Authority Permitting Other Than Full and Open

Competition.

The statutory authority permitting other than full an open competition is 41 U.S.C. 3304(a)(1) implemented by the FAR Subpart 6.103-1 entitled “Only One Responsible Source and No Other Supplies or Services Will Satisfy Agency Requirements.”

5. Demonstration that the proposed contractor’s unique qualifications or the nature of the acquisition requires use of the authority cited.

(a) Rationale for using the authority cited in section 4 of the J&A;

As advances in X-ray, CT, and millimeter wave-based on-person scanning systems, known as Advanced Imaging Technology (AIT) systems, have been made, manufacturers of these screening systems must develop new threat detection algorithms on the image data sets generated by these screening systems. Building and testing the detection algorithms requires access to large amounts of high-quality, labeled, and curated data. For homeland security applications, there are several barriers to getting this data: 1) threats are constantly evolving;

2) tailored sensor image data (particularly images containing threat data) is time consuming and expensive to acquire; 3) manual image annotation and labeling is time consuming and expensive to do; and 4) security datasets are massive, difficult to share, and security sensitive information restricted. This procurement continues the maturation of a tool that would produce synthetic raw X-ray and millimeter wave image data from a prompt that screening system manufacturers, as well as Government testers, could use to develop their detection algorithms, thereby reducing the resources otherwise required. The development of a capability to synthetically generate raw image data, that is fundamentally computationally same as the raw image data produced by CT and AIT systems, will save resources and time for developers of CT or AIT systems. Cignal is the only company that has developed this capability and periodic demonstrations to S&T indicate that the solution will conform with data format and other attributes required for successful detection algorithm development.

(b) Details covering what events lead to the situation requiring use of other than full and open competition procedures including whether any portion of the work can be segregated to allow for competition;

The contractor Cignal was the successful offeror to the solicitation 70RSAT18R00000024 in 3QFY19 under the title “Object Recognition and Adaptive Algorithms in Passenger Property Screening.” This initial Silicon Valley Innovation Program (SVIP) Other Transaction Agreement (OTA) “Phase I” award resulted in satisfactory demonstration of the development of their novel synthetic image generation. The contractor was awarded Phases 2, 3, and 4, with each award dependent upon demonstration of satisfactory development and maturation.

The synthetic data image generation solution is currently at TRL 4 for CT/X-ray images and TRL 3 for millimeter wave images. The follow-on work required by the Program includes physics-based scene modeling and coding to support physical reality modeling and coding and creating corresponding deep learning models to further their novel proprietary and software architecture and unique patented approach for generating both X-ray and millimeter wave images from a single base image generator. Due to the proprietary and patented novel architecture and deep learning coding, another contractor would not have access to the software details, deep learning structures, mathematical equations, or architecture details necessary to be able to carry the work further. No portion of the work can be segregated to allow for competition; all portions of the requirement work in concert towards the goal of technology maturation.

(c) Summarize alternatives considered and why they will not work;

In the course of closely monitoring the satisfactory development by the contractor, no other company has been discovered to have developed a similar capability. The S&T Transportation Security Laboratory (TSL) held a Data Consortium Industry Day in February 2026, with “synthetic data” a topic for discussion. The Industry Day lead informed the S&T Program that only two other companies expressed interest in, but demonstrated no capability in, the field of synthetic data. The ability to generate synthetic image data for algorithm development or modification is anticipated to save considerable time and resources by both the private sector and Government and contribute to the detection performance of checkpoint screening systems, thereby supporting the mission to promote public safety. Alternative approaches to generating synthetic images focus on human discernment training but fail due to higher false alarm and negative rates when applied to machine vision training because computers process images at much higher bit precision than the human eye can perceive.. This project addresses that gap, which requires a significant amount of physics-based scene modeling and mathematical equations to match the synthetic data with real data with high bit precision.

(d) Impact to the mission that would result if the J&A is not approved and, consequently, the product or service not provided.

If this J&A is not approved, DHS will continue spending many millions of dollars and many months for creating real image data sets for each emerging threat with no viable solution in sight except for this one.

• If using “Only One Responsible Source and No Other Supplies or Services Will

Satisfy Agency Requirements” as your authority, also include the following:

o Details on the specific requirements (not what the equipment, process, or service is capable of doing);

Cignal will develop and demonstrate software tools that support synthetic data generation suitable for public algorithm development (non-sensitive) and Government development, testing, and evaluation (sensitive). Cignal will collaborate with potential end users at S&T, TSA, and elsewhere to validate tools and generated images for suitability for machine learning development, training, testing, and evaluation over the course of the contract.

o Information on why only one contractor is capable of fulfilling the requirements;

The Program wishes to advance the solution developed by Cignal because it is a unique solution that promises to provide a prompt-based synthetic data image generation capability, and market research showed that no other vendor has a capability that approaches the solution developed by Cignal. Cignal has advanced the solution to generate an initial set of synthetic images for CT/X-ray systems and has recently made that dataset available to the public, partly to elicit feedback from potential future users. Further development of the CT/X-ray aspect of the solution is required by the Program, as well as the maturation of the same synthetic image generation ability for millimeter-wave based screening systems. The overall aim of the Program is to realize a solution for synthetic data image generation that meets the same technical specifications that real images would provide. Cignal’s novel machine learning architecture image generator and their unique, cost-effective concept for different deep learning “heads” for tailoring the generator’s output for different modalities that address the gap between human vision and machine vision discernment cannot be duplicated by, let alone handed-off to any other performer since Cignal retains all rights to the data, patents, trademarks, copyrights, and intellectual property afforded to them under the previous SVIP OTAs. Further no one else has demonstrated ability to address this human-machine simulation gap for X-ray and millimeter wave data sets for use by Original Equipment Manufacturers. The contractor has demonstrated their novel and proprietary patented approach through several years of S&T funding under SVIPand released the first prototype data set that demonstrated the efficacy of the technology to Government stakeholders confirming the value of proceeding with further maturing the technology.

o Why other proposed contractors are unable to meet the requirements;

There are no other potential contractors to which this work can be handed off for further maturation because of the significant learning curve, proprietary and patented nature of the software architecture, and uniqueness and cost effective multimodal approach for generating X-ray and millimeter wave images. The Program maintains awareness of developments in the synthetic data realm but has neither been approached by any other potential contractors nor aware of any addressing the gap between vision and machine discernment.

o Unique capabilities or qualifications of the contractor that form the basis for the justification;

Cignal successfully competed against the original topic call with their unique and cleverly efficient, multi-modal approach that has advanced the capability to the point of demonstrating their ability to switch “AI heads” while using the same basic primary image generator, which makes it easier to produce synthetic image data sets for other kind of sensor image sets beyond just X-ray CT and millimeter wave AIT that TSA is interested in including radio frequency and terahertz imaging systems. It meets technical aspects of the requirement. The capability prototype was evaluated by S&T and other Component stakeholders which confirmed that Cignal’s progress and capability not only more than meets TSA requirements for X-ray CT and millimeter-wave AIT data sets but also for other systems as well, including scanners based on radio-frequency and terahertz imaging and for other scanner configurations including hand-held on-person scanners and miniature X-ray CT systems. The delivered image data and evaluations also validated Cignal’s patented concepts and proprietary architecture and software approaches and their experience and capability to further this work to commercialization.

o Information on why and description of the extent to which the Government will be harmed if the supply or service is not acquired from that vendor.

Screening systems at checkpoints require an ATR algorithm that identifies contraband on a person or in a bag. These algorithms must be trained with real image data from CT/X-ray or millimeter wave-based systems. That image data is obtained by both an original equipment manufacturer and the Government by building hundreds of packed bags with and without contraband and recruiting humans to serve as mock passengers with and without contraband. The more such images can be collected (thousands), the better the resulting detection algorithm will be. The ability to produce synthetic data via a prompt will save a considerable amount of time and resources required by making the synthetic images widely available for algorithm development. In addition, at times an emerging threat becomes known, DHS asks the manufacturers of deployed checkpoint systems to modify their ATR algorithm to include the ‘new’ threat.

The ability to leverage synthetic images will significantly reduce the time required for deployed systems to include a new target in the ATR algorithm. As threats evolve and air travel increases, deployed systems are stretched to their limits.

Synthetic data will help to stretch the abilities of those deployed systems and reduce inefficiencies in algorithm development. The longer this technology is delayed, the Government will continue to expend significant amounts of resources for real data set collection (many millions of dollars every year) and long deployment times where a new threat could slip through risking the security of passenger travel in the meantime.

6. Description of Efforts Made to Ensure that Offers are Solicited from as Many Potential

Sources as is Practicable.

A synopsis will be posted on SAM.gov the week of August 24, 2026. Any potential sources will be addressed accordingly.

7. Determination by the Contracting Officer that the Anticipated Cost to the Government will be Fair and Reasonable.

The Contracting Officer determines that the anticipated price will be fair and reasonable based on prior acquisition history from the previous phases of development.

8. Description of Market Research.

(a) Describe the market research techniques utilized, the specific results of the market research, the date when the market research was conducted, and how the market research was used by the contracting officer to determine price reasonableness.

The initial award was solicited on SAM.gov (70RSAT18R00000024) on April 26, 2019, under the title “Object Recognition and Adaptive Algorithms in Passenger Property Screening.” Three companies sent in proposals and Cignal LLC was successful. This was treated as Phase I of the S&T Silicon Valley Innovation Program (SVIP) Other Transaction Authority, with a total of five phases possible if the contractor was successful in prior phases. Cignal was successful in each phase. In 2025, the ability to form OTA agreements was taken away from DHS, otherwise the Program planned to award a Phase 5 for maturation and transition steps. Due to the progressive stages of this effort, the Program has not conducted market research through traditional means. Instead, market research has involved the Program being receptive to outreach by potential performers, directly and through evaluation of proposals via the LRBAA portal, which resulted from several industry and public engagement events that included presentations by Cignal. The SVIP program hosted an event in 2024 that included showcasing Cignal’s emerging solution. S&T issued a press release in 3QFY26 describing the availability of data sets for algorithm developers. In February 2026, S&T TSL held an industry day on creating a data consortium, with the topic of synthetic data being secondary. The Government SME was interviewed for this market research and relayed that only two other companies expressed interest in entering the synthetic data domain but did not display or demonstrate any actual capabilities gained to date.

(b) All rights to data, intellectual property, patents, trademarks, and copyrights were retained by Cignal under the terms and agreements in the OTA awards made under the SVIP program. The Government retained unlimited rights only to the deliverables. It is not possible for any other entity but Cignal to continue this work given these retained rights.

9. Any Other Facts Supporting the Use of Other Than Full and Open Competition.

Not applicable.

10. A Listing of the Sources, if Any That Expressed, in Writing, an Interest in the

Acquisition.

A synopsis was issued on SAM.gov the week of August 24, 2026. All respondents were addressed accordingly.

11. A Statement of the Actions, if Any, the Agency May Take to Remove or Overcome Any Barriers to Competition Before Any Subsequent Acquisition for Supplies or Services Required.

In general, the S&T Program anticipates future solicitations for methods to increase efficiency and reduce time and resources required for threat detection algorithm development, but the ability to have Cignal’s synthetic image data generation solution will first need to be given to end-users and explored for next steps in this field. The increasing dominance of artificial intelligence and machine learning suggest that the landscape of algorithm development is likely to change over the next several years but at present, technology has not advanced broadly enough in this field to warrant specific planning.

(Remainder of this page left blank intentionally)

12. Contracting Officer’s Certification. I certify that the data supporting the recommended use of other than full and open competition is accurate and complete to the best of my knowledge and belief.

/s/ Contracting Officer

13. Technical/Requirements Personnel Certification. I certify this requirement meets the

Government’s minimum need and that the supporting data, which forms a basis for this justification, is complete and accurate.

/s/ Technical Representative

APPROVAL:

/s/ Procuring Activity Advocate for Competition