Prove You're 18 Without Showing Who You Are: The Cryptography Big Tech Won't Use
Source: Dev.to
Engineering a ‘Yes/No’ without the ‘Who/Where’
Traditional facial comparison
In the world of facial comparison—the core technology leveraged at CaraComp—the standard output is usually a similarity score derived from Euclidean‑distance analysis. Two face prints are mapped into a high‑dimensional vector space, and the distance between them is calculated. In a traditional verification flow, that vector (the embedding) is highly sensitive biometric data. Storing it creates a biometric honeypot.
Privacy‑preserving age checks
Recent privacy‑preserving age‑check proposals suggest a new architectural pattern: Zero‑Knowledge Proofs (ZKPs) on the edge. Instead of sending the image or the vector to the server, the computer‑vision model’s output becomes an input for an arithmetic circuit.
For developers, this means the typical POST /verify endpoint—often returning a JSON object filled with personally identifiable information (PII)—can be replaced by a workflow where the server receives only a proof that is computationally impossible to reverse‑engineer into a face print or a birthdate.
New workflow
- The client‑side CV model performs the Euclidean‑distance analysis.
- The result is fed into a ZK‑SNARK (Zero‑Knowledge Succinct Non‑Interactive Argument of Knowledge).
- The server receives a proof that the check passed, without ever seeing the raw biometric data.
Comparison vs. recognition
The news commentary often conflates facial comparison with facial recognition, but for developers the difference lies in database architecture:
- Facial recognition requires a 1:N search against a gallery, which is computationally expensive and privacy‑invasive.
- Facial comparison (1:1) is what is needed for verification.
By keeping the scope to comparison and layering it with ZKPs, the need for a “Big Brother” infrastructure is eliminated. There is growing demand for enterprise‑grade Euclidean analysis that is affordable (e.g., $29 /month) without sacrificing security.
Performance challenges
Implementing these arithmetic circuits without degrading performance on mobile hardware is a key challenge for the community. Bridging the gap between heavy OpenCV/TensorFlow processes and the rigorous requirements of cryptographic proof generation is essential.
Future considerations
As we build these tools, we must ask: Are we building a gate or a tracker? If a verification stack requires storing a face print to prove a user is 18, it is not an age check—it is a surveillance node.
The future of biometric engineering is not just about higher accuracy metrics; it is about proof‑of‑attribute without data leakage. When designing verification flows, consider whether you prioritize the completeness of the user profile or actively move toward a Zero‑Data architecture.