Face Authentication: Template-Free, Biometric-Blind Login

Face authentication is a passwordless sign-in method that confirms who someone is by checking their live face against an enrolled reference, rather than asking for a password or a one-time code. Instead of proving you know a secret, you prove you are the person — captured live, in the moment, at the point of login.

It is worth separating two things that sound alike. Face authentication verifies a claimed identity at login (one-to-one: "is this the enrolled person?"), which is what SenseCrypt does. Facial recognition login is the everyday phrase people use for the same experience — you look at the camera and you are in. What matters for security and privacy is not the phrase but what the system keeps behind the scenes: does it store your face, or does it store something that can never be turned back into your face?

SenseCrypt, built by Seventh Sense (a Singapore deep-tech company), answers that question the strong way. Users enroll from a photo on file and sign in by face; the face match runs on the device; and the only thing SenseCrypt retains is a sealed, unlinkable, non-reversible face token that contains no PII and that even Seventh Sense cannot reverse. That is what "template-free" and "biometric-blind" mean.

At enrollment, SenseCrypt turns the reference photo into a sealed, non-reversible face token — not a stored image and not a matchable biometric template. At sign-in, the device captures a live face, checks that it belongs to a real, present person (liveness / presentation-attack detection), performs the face match on the device, and returns a cryptographic proof rather than the biometric itself.

That flow rests on two independent guarantees working together. A real FIDO2/WebAuthn passkey (ES256) proves the device through the browser's origin binding, and the live-face check proves the person. Neither alone is the whole answer: the passkey stops credential phishing, and the face check stops a stolen or shared device from signing in as someone else.

No. SenseCrypt stores no face image and no biometric template. The only thing it retains is a sealed, unlinkable, non-reversible face token that contains no PII and that even Seventh Sense cannot reverse. Because matching runs on the device, your raw face data is never centralized in the first place.

This is the whole point of template-free, biometric-blind design. Many face-based systems store a biometric template — a mathematical representation of your face that can be matched, and in some designs reversed or correlated across services. A store like that is a permanent liability: a face is not a password you can reset after a breach. SenseCrypt removes the template entirely, so there is simply no biometric honeypot to steal, subpoena, or correlate.

The claim is not merely asserted. SenseCrypt's face recognition is evaluated in the NIST FRTE (Face Recognition Technology Evaluation) under developer ID seventhsense-000, in the program since 2021, and its liveness / presentation-attack detection holds iBeta ISO 30107-3 certification at Levels 1 and 2 — independent, third-party validation rather than a self-graded marketing claim.

The defense against deepfakes, printed photos, replayed video, and masks is liveness detection — also called presentation-attack detection (PAD). Its job is to confirm the face at the camera belongs to a real, present person rather than an artifact. SenseCrypt's liveness holds iBeta ISO 30107-3 certification at Levels 1 and 2, the recognized independent benchmark for how well a system resists presentation attacks.

Recognition accuracy and spoof resistance are two separate questions, and SenseCrypt is measured on both. NIST FRTE evaluates how accurately the face recognition matches the right person and rejects the wrong one; iBeta ISO 30107-3 evaluates how well the liveness layer rejects presentation attacks. Passing one does not imply the other, which is why SenseCrypt is independently tested on each.

Security is layered rather than resting on the face alone. Even a perfect spoof would still have to defeat the FIDO2/WebAuthn passkey bound to a legitimate device, and phishing-resistant origin binding means an attacker cannot simply relay the login through a fake site. The attacker needs the right device and a live person who passes liveness — a materially higher bar than stealing a password or intercepting a one-time code.

The axis that separates SenseCrypt from a typical face-authentication vendor is not whether it uses a camera — everyone does — but what it stores afterward. Many server-side face systems keep a biometric template; SenseCrypt keeps only a sealed, non-reversible token and does the match on the device. The table below compares on that neutral axis rather than on any single competitor's internals.

Read the row "Stores a biometric template or face image?" first. SenseCrypt's answer is a clean No for a server-side face IdP, and that single design choice cascades into the rest: no central biometric database to breach, nothing reversible to correlate across services, and a straightforward answer when a regulator or customer asks "do you hold a biometric?"

No — SenseCrypt's face authentication is not your phone's Face ID. Face ID is Apple's on-device screen unlock, tied to one device's secure enclave. SenseCrypt runs its own template-free face check: matching happens on the device, but it is SenseCrypt's independently evaluated face recognition (NIST FRTE) and liveness (iBeta ISO 30107-3), not the phone's built-in biometric, that decides whether the enrolled person is present.

This distinction matters for two reasons. First, portability: because SenseCrypt's face check is its own, the assurance of "the right person" travels with the account rather than being pinned to a single handset's sensor. Second, proof: SenseCrypt's recognition and liveness are independently certified, so the security posture does not depend on whatever biometric a given device happens to ship.

To be clear about the passkey layer: SenseCrypt does use standard FIDO2/WebAuthn passkeys (ES256), and those passkeys work exactly the way passkeys are meant to — a device platform authenticator can guard the key. SenseCrypt embraces device security on the passkey path; it simply does not outsource the who-is-this decision to a device's built-in face unlock. The passkey proves the device; SenseCrypt's own live-face check proves the person.

SenseCrypt is a standards-native identity provider, so face authentication drops into an existing estate rather than forcing a rebuild. It speaks the protocols identity teams already use, which means face login can become an option inside your current SSO and provisioning setup instead of a parallel silo.

That standards footing is what lets template-free face authentication be added as a federated login path, a step-up factor for high-assurance actions, or a primary IdP for customer identity — without giving up governance, provisioning, or auditability.

• Protocols: OIDC, OAuth 2.0 (with PKCE and PAR), SAML 2.0, SCIM 2.0, and CIBA.
• Enterprise controls: RBAC, audit logs, and multi-tenant isolation.
• Authentication core: FIDO2/WebAuthn passkeys (ES256) plus an on-device live-face check.
• Pricing: a flat $1 per user per month, with a 30-day free trial and no card required.