Mirage
Noshaba Cheema, Amir Razagh, Vlada Kandyba · Team Mirage
Submitted to Defensive Acceleration Hackathon. Sprint projects are early-stage work by participants, not Apart Research publications.
The deepfake detection market is rapidly growing in response to the explosion of AI-generated fake content. In 2023 the global deepfake detection market was valued at only about $213 million, but it is forecast to surge to roughly $3.46 billion by 2031 – a 40%+ CAGR growth trajectory. This explosive growth is driven by the rising threat of deepfake fraud and misinformation across industries. For example, identity fraud attacks using deepfakes jumped 31× from 2022 to 2023 in one report, underscoring the urgent need for reliable detection. Major sectors are already seeking solutions: banking and financial services want to prevent deepfake identity/document fraud; media and entertainment companies need tools to preserve trust and authenticity by stopping the spread of fake videos and images; governments and defense agencies view deepfakes as a national security risk; and social media platforms are exploring anti-deepfake tech to maintain a safe environment for users. Journalists and fact-checkers in particular are looking for ways to verify images, videos, or audio clips, as public trust in media is eroded by convincing fake visuals. This is where Mirage fits in: it targets a clear pain-point across these domains by offering a multi-modal deepfake detection agent (covering text, images, and video) that delivers a “genuineness score” for content. Its use of blockchain (NEAR Protocol) for immutable logging of verification results is a unique differentiator, adding transparency and public trust to the detection (similar in spirit to initiatives like Project Origin which use tamper-proof ledgers to certify media authenticity). Likewise, leveraging a TEE (Trusted Execution Environment) for the AI agent addresses security and privacy concerns, ensuring sensitive analyses are done in a private, verifiable manner. Overall, Mirage is well-positioned to serve its target audiences – enterprises, social media platforms, and journalists – by filling the market need for a trustworthy, real-time deepfake detection solution that is both highly accurate and credibly transparent (thanks to on-chain verification).
Reviews
No public critique yet.
Cite this project
@misc{cheema2025mirage,
title = {{Mirage}},
author = {Noshaba Cheema and Amir Razagh and Vlada Kandyba},
year = {2025},
month = nov,
note = {Submitted to Defensive Acceleration Hackathon, an Apart Research Sprint},
howpublished = {\url{https://apartresearch.com/sprints/projects/mirage-evhs}},
url = {https://apartresearch.com/sprints/projects/mirage-evhs}
}More from Defensive Acceleration Hackathon
- View project: Neops - DevSecOps for the AI era
Neops - DevSecOps for the AI era
Broad Bros
NEOps is a CLI-based tool that embeds AI safety into your product lifecycle from day one. While development teams routinely build cybersecurity checks, AI-safety often comes later—or not at all. NEOps fills that gap by …
- View project: Assisted Audit of Solana Programs
Assisted Audit of Solana Programs
GLAM
Multi-agent solution that assists in auditing Solana programs, allows to consolidate audit findings into a knowledge base, and can integrate into CI/CD pipelines to prevent security regressions.
- View project: Mechanistic Watchdog
Mechanistic Watchdog
SL5
Mechanistic Watchdog is a mechanistic-interpretability-based “cognitive kill switch” for language models. Instead of only filtering final text, we monitor a model’s internal activations in real time and learn linear …