Skip to content
VULT SecurityGet started

Your Bank Knows the Phone Unlocked. It Does Not Know You Approved the Transfer.

Why a successful device check is not enough to prove which human approved a transaction, and how VULT Security connects the verified person to the exact action.

Device biometrics can show that a phone approved an action, but not which verified human did. VULT Security verifies the human is present and binds their approval to the exact event.

Your bank asks for a Face ID.

Your phone approves it.

The transfer goes through.

Everyone behaves as though the bank just proved you were there.

But that is not what happened.

The phone determined that its current authentication requirements had been satisfied. The bank received a successful response and treated that response as proof of the account holder.

Those are not the same thing.

Device-bound biometrics can be excellent at protecting a phone or releasing a credential stored on it. What they do not independently prove is that the person behind the device is the government-verified owner of the bank account, payment account, or identity being represented.

Yet across banking, payments, money movement, and digital commerce, we often treat device approval as human proof.

That gap is where fraud lives.

What Does a Successful Device Biometric Actually Prove?

A successful device biometric generally establishes that the device accepted a face, fingerprint, or other factor enrolled within its local security system.

The result may unlock the phone, open an application, or authorize use of a device-held key. The outside organization usually receives a success signal. It does not receive the biometric and does not independently compare the person against a verified legal identity.

Apple describes this boundary directly. Applications using Face ID or Touch ID are notified whether authentication succeeded, but they cannot access the biometric information associated with the enrolled user. Apple Platform Security

That is strong privacy architecture for a local authenticator.

It also reveals the limitation.

The bank learns that the device said yes.

The bank does not necessarily learn which verified human caused it to say yes.

We Mistake Device Possession for Human Identity

The industry has gradually collapsed several different ideas into one:

  • The person has access to the registered phone
  • The phone accepted its local authentication method
  • The rightful account holder is using the phone
  • The authorized human understands and approves the transaction

Only the first two may have been technically established.

The last two are often assumptions.

This matters because a device can change hands. Passcodes can become known. Accounts can be recovered. Sessions can remain active. Credentials can synchronize. Devices can be compromised. Local biometrics can be re-enrolled when someone gains sufficient control of the device.

Suppose someone steals a phone and also obtains its passcode through observation, coercion, social engineering, or another compromise. Depending on the device and application controls, that person may be able to add their own biometric, regain access through a fallback path, or continue using an already authenticated session.

The next biometric check may work exactly as designed.

The device sees the newly enrolled face or fingerprint and approves it.

To the bank, the signal may look the same: authentication successful.

But the human is different.

Some devices and financial applications detect biometric enrollment changes, invalidate keys, require reauthentication, or impose additional controls. Those safeguards matter, and they should be used. They are not universal, and they still depend on the surrounding device, recovery, enrollment, and account-security chain.

The problem is not that the biometric necessarily failed.

The problem is that we asked the device a question it was never designed to answer:

Is the authorized human present and personally approving this exact action?

AI Has Made the Assumption More Dangerous

AI did not suddenly make every password, passkey, or device biometric useless.

It did something more subtle and, in many ways, more dangerous. It weakened our confidence in the human-facing systems surrounding them.

AI can generate convincing emails, text messages, documents, voices, photographs, and video. It can make social engineering faster, cheaper, more personal, and easier to scale. When a phone number, email account, text channel, help desk, or recovery process is compromised, AI-generated impersonation can make the attacker appear far more credible inside that channel.

A familiar voice is no longer reliable proof.

A polished email is not proof.

A realistic video is not proof.

A text arriving from the expected conversation is not proof that the rightful person wrote it.

Even the right device is not proof that the rightful human is holding it.

The signals we once used to infer identity can now be manufactured, manipulated, or presented by someone else.

That is the Human Substitution Problem: information, credentials, devices, and lines of code are allowed to stand in for the human.

Device Biometrics Are Useful. They Are Not Identity Proof.

This is not an attack on Face ID, Touch ID, Android biometrics, secure hardware, or passkeys.

They are useful technologies. In many cases, they are far better than passwords alone.

Apple states that Face ID data remains on the device and is encrypted within the Secure Enclave. That is a meaningful privacy and security benefit. Apple Face ID Security

The limitation is not necessarily how well the device recognizes an enrolled user.

The limitation is what that recognition means to an outside organization.

Unless the biometric enrollment was independently bound to validated government identity evidence and the organization can rely on that binding, the device does not establish which legal person enrolled the biometric.

The National Institute of Standards and Technology makes a related distinction. NIST states that a biometric characteristic is not an authenticator by itself. In common device-based systems, the biometric acts as an activation factor for something the user possesses, such as a device-held cryptographic authenticator. NIST SP 800-63B

In plain language, the biometric often activates the device's authority.

It does not independently prove the human's identity.

VULT Security Starts With the Human

VULT Security approaches the problem from the opposite direction.

The device is not treated as the person. It is used as a secure capture mechanism through which the person can prove themselves.

During VULTid enrollment, VULT Security validates a government-issued identity document. The live human is biometrically matched to the portrait belonging to that identity. Real-time liveness detection confirms that a live person is participating rather than a photograph, replay, recording, or previously captured biometric.

This establishes a reusable verified identity bound to its rightful human owner.

When a bank, payment platform, enterprise, or government system needs that person to authorize a protected action, VULT Security performs a new server-side biometric Proof-of-Presence check.

The question is no longer whether a familiar device unlocked.

The question becomes whether the verified human is present now.

For an authorization event, VULT Security goes further by binding that human's authenticated approval to the exact action presented.

The amount. The destination. The account. The instruction. The moment.

Only that human. Only that event.

Why Server-Side Verification Provides Stronger Human Assurance

Device-side biometric matching asks the device to evaluate its locally enrolled user and release a device-bound capability.

Server-side matching allows VULT Security to compare a new live capture against the protected template belonging to a government-verified VULTid.

That difference creates three important advantages.

First, identity is established independently of the current device. Replacing, stealing, sharing, or reconfiguring a phone does not replace the verified human.

Second, the organization receives proof tied to a government-verified identity rather than only a local authentication-success signal.

Third, VULT Security can bind the verified human's presence and explicit approval to the exact event the organization presented.

This does not make every form of centralized biometric storage safe. A conventional central database containing ID images, selfies, and reusable biometric records can become an extraordinarily valuable target.

NIST recognizes that central biometric comparison creates greater potential for attacks at scale and therefore requires additional protections for central verifiers. NIST SP 800-63B

VULT Security addresses that risk by separating server-side human assurance from the retention of raw identity artifacts.

VULT Security Does Not Keep the ID or Face Image

VULT Security uses a government-issued ID and a live facial capture to establish the verified identity.

After verification, VULT Security does not retain the government ID image or the raw facial image.

VULT Security retains a protected, non-reversible facial template.

The distinction is important.

A government ID image can be read, copied, and visually inspected. A face photograph can be viewed and reproduced. A non-reversible facial template is a mathematical representation designed for biometric comparison, not a stored photograph of the person.

It is designed so the original face image cannot be reconstructed from the template.

VULT Security protects that template with homomorphic encryption and rotating encryption keys.

Homomorphic encryption allows supported computations to be performed on encrypted information without first exposing the underlying data in readable form. NIST describes homomorphic encryption as a special form of encryption that supports evaluation over encrypted data. NIST Fully Homomorphic Encryption

Rotating keys add another layer of protection by changing the keys used to secure stored templates rather than relying indefinitely on one static key.

The architecture is intentionally narrow:

  • The government ID establishes the verified identity but is not retained as an ID image
  • The live face establishes the rightful human but is not retained as a face image
  • The organization never receives or maintains either artifact
  • VULT Security retains only the non-reversible facial template needed for future comparison
  • The template remains protected through homomorphic encryption
  • Encryption keys rotate rather than remaining permanently static

This is server-side verification without creating a conventional database of ID cards and face photographs.

The Difference in One Table

Question

Device-Bound Biometrics

VULT Security

What is evaluated?

A biometric enrolled on the device

A live biometric compared with a government-verified VULTid

Who performs the comparison?

The local device

VULT Security server-side verification

What does the organization receive?

A device or authenticator success response

Verified-human presence and, when required, event-specific approval

Is legal identity independently established?

Not necessarily

Yes, through validated government-issued identity evidence

Is the approval bound to the exact external event?

Not inherently

Yes

Does the organization store the ID or biometric?

Usually not through the device biometric API

No

Does VULT Security retain the raw ID or face image?

Not applicable

No

What persists with VULT Security?

Not applicable

A non-reversible facial template protected by homomorphic encryption and rotating keys

Authentication Is Not the Same as Authorization

Identity verification establishes that a person matches a claimed identity.

Authentication checks whether the required proof has been presented.

Traditional authorization determines whether an account has permission to perform an action.

VULT Security answers the remaining question:

Is the authorized human present and personally approving this exact action?

The authorized human is the person who already has permission or authority to approve the action. VULT Security does not grant that authority. VULT Security confirms that the person who holds it is actually present and giving approval.

That is Event-and-Identity-Bound Authorization.

Verified Identity. Human Presence. Authorization Event. Authenticated Approval.

One private moment, backed by a protected evidence record.

Permission allows the action. VULT Security proves the authorized human approved it.

The Device Should Capture the Human, Not Substitute for Them

Banks and payment platforms do not need to abandon device biometrics.

They need to stop treating them as proof they were never designed to provide.

Use the phone for capture.

Use the device for communication.

Use secure hardware to protect local secrets.

But when money, rights, access, safety, or accountability are at stake, prove the human.

VULT Security does not ask whether the device recognizes someone.

VULT Security proves whether the verified human is present and personally approving the exact action.

The device still matters.

It just no longer gets to stand in for the person.


Frequently Asked Questions

Do device biometrics prove the identity of the person using a banking application?

Not necessarily. Device biometrics generally establish that the device accepted a locally enrolled biometric or authentication path. Unless that enrollment was independently tied to validated identity evidence, the result does not prove a particular government identity to the bank.

Can someone who steals a phone add their own biometric?

Potentially, if that person also gains sufficient control of the device, such as its passcode or account-recovery path. Security behavior varies by device and application. Some financial applications detect biometric enrollment changes or require new authentication, but organizations should not assume every implementation prevents this scenario.

Has AI defeated device biometrics?

Not categorically. AI has made impersonation, phishing, synthetic communications, and attacks against human-facing recovery processes more convincing and scalable. Device biometrics remain useful, but they do not resolve the underlying question of which government-verified human is acting.

Why does VULT Security perform biometric verification server-side?

Server-side verification allows VULT Security to compare real-time biometric presence against the protected template belonging to a government-verified VULTid. The decision does not depend solely on the biometric currently enrolled on a particular device.

Does VULT Security store government IDs or face photographs?

No. VULT Security uses the ID and live facial capture during verification but does not retain the government ID image or raw face image after verification is completed.

What biometric information does VULT Security retain?

VULT Security retains a non-reversible facial template used for future biometric comparison. It is protected using homomorphic encryption with rotating keys.

What is real-time Proof-of-Presence?

Real-time biometric Proof-of-Presence confirms that the human bound to a verified identity is physically present when verification or approval is required.

What does VULT Security prove that a device biometric does not?

VULT Security proves that the live human originally matched to a validated government-issued ID is present for the protected event. When approval is required, VULT Security also binds that verified human's authenticated approval to the exact action presented.

Sources

  1. Apple Platform Security: Uses for Optic ID, Face ID and Touch IDWhat applications receive from local biometric authentication and why they cannot access the enrolled biometric data. Apple Support (NZ)
  2. Apple: About Face ID Advanced TechnologyOn-device facial matching, Secure Enclave protection and biometric privacy. Apple Support
  3. Apple: About Stolen Device Protection for iPhoneAdditional biometric checks and security delays for sensitive changes, including adding or removing enrolled biometrics. support.apple.com
  4. Android Developers: Biometric Enrollment and Key InvalidationHow applications can invalidate authentication keys when biometric enrollment changes. See setInvalidatedByBiometricEnrollment. developer.android.com
  5. NIST SP 800-63B-4: Authentication and Authenticator ManagementBiometric activation factors, local versus central comparison, and protections required for centrally processed biometrics. pages.nist.gov
  6. NIST SP 800-63A-4: Identity Proofing and EnrollmentValidating identity evidence, matching the applicant to that evidence, and presentation-attack detection requirements. pages.nist.gov
  7. W3C: Web Authentication, User VerificationThe distinction between verifying an authenticator’s user and identifying a particular real-world person. www.w3.org
  8. FBI IC3: Criminals Use Generative Artificial Intelligence to Facilitate Financial FraudCriminal use of generated text, images, cloned voices and synthetic video for impersonation and fraud. ic3.gov
  9. NIST: Fully Homomorphic EncryptionHow homomorphic encryption supports computation on encrypted information. CSRC
  10. OWASP: Transaction Authorization Cheat SheetDistinguishing authentication from transaction authorization and enforcing authorization for each specific operation. cheatsheetseries.owasp.org
  11. VULT Security: Verified-Human Identity and Event-Based AuthorizationVULT Security’s stated capabilities for government-ID-linked identity, server-side verification, real-time presence and protected authorization evidence.