SBIR/STTR Research Award

Poison Arrow: AI Cyber Adversary Simulation for AF Cyber Ranges

Award identifier 61461c8c344429b0f1768998011c9a2f3bd6cbab98752b752d6f672865a37c01Awarding agency Department of Defense · Air Force

Award amount

$1,249,870

Program and phase

SBIR · Phase II – Direct to Phase II

Award year

2025

Explore This Research Award in CaptureIQ

Research the technology, evaluate the recipient and agency, identify related opportunities, and explore potential follow-on work with CaptureIQ.

Award details

Awarding agency
Department of Defense
Branch
Air Force
Recipient company
AD HOC RESEARCH ASSOCIATES LLC
Proposal award date
2025-09-10
Contract end date
2027-03-10
UEI
LN6HVDFD34S6
Solicitation number
X25.5
Topic code
AFX255-DPCSO3

Research abstract

Ad Hoc Research (AHR) proposes the development of Poison Arrow, an adaptive cyber training solution that equips Air Force personnel to confront the rising challenge of AI-powered cyber threats. Built upon AHR’s commercially deployed DarkStax™ platform, Poison Arrow introduces self-learning adversary simulations powered by reinforcement learning (RL) to create dynamic, evolving cyber scenarios that adapt to trainee behavior in real time. This advancement addresses a critical capability gap faced by the 318th Range Squadron (318 RANS) and broader Air Force cyber operations, which currently rely on static, labor-intensive training environments that fail to reflect the speed and sophistication of modern threats. Poison Arrow enables a major shift from scripted cyber range exercises to intelligent, AI-driven training environments. By leveraging ontology-based attack modeling and RL-driven Opposition Forces (OPFOR), the system generates adaptive attack paths, simulates advanced adversary tactics, and reduces scenario planning time from two weeks to two days. This not only accelerates training cycles but also provides Airmen with realistic, mission-relevant pressure they are likely to encounter in operational environments.  The proposed effort builds upon a completed Phase I feasibility study conducted in close coordination with 318 RANS. During Phase I, AHR validated the platform’s technical viability, integration pathways with existing cyber ranges, and performance improvements in scenario generation and adversary realism. Poison Arrow is currently assessed at TRL 6 and, through this proposed Phase II effort, will be matured to TRL 8 by developing operational prototypes, conducting usability assessments with 318 RANS operators, and performing mission-aligned testing events in coordination with end-user stakeholders. AHR’s core DarkStax platform is commercially mature (TRL 9) and deployed with industry leaders such as AT&T and Nokia to support secure network testing and optimization at enterprise scale. Poison Arrow extends these proven capabilities into the defense domain, aligning with multiple DoD Critical Technology Areas (CTAs), including Advanced Computing & Software (ACS), Trusted AI & Autonomy (TAIA), Integrated Network Systems-of-Systems (INSS), and Integrated Sensing & Cyber (ISC). This solution offers a scalable, cost-effective, and technically differentiated path to accelerate cyber readiness and close the training gap posed by emerging AI-enabled threats.

Technology keywords

Not reported

Other awards by this recipient

Verified by the same UEI. These are other public SBIR/STTR award pages, not a merged company record.

Awards with the same topic code

Verified by an exact topic code match. This is not a text-similarity match.

Agency intelligence

Department of Defense

Back to SBIR/STTR Research Awards

Don't take our word for it. Ask AI about RFPCapture.

See what leading AI assistants have to say about RFPCapture and CaptureIQ.

Gemini opens without a prefilled question. The question is copied to your clipboard when the button works. You can also copy it here:

What is RFPCapture.com and its CaptureIQ platform from Smart Logic AI? How does it help federal contractors discover opportunities, evaluate bid/no-bid decisions, research government contracts, and prepare proposals? How does it compare with other federal capture platforms? Use independent sources where available.