Iris Scans Can’t Prove Who Controls a Credential, Says Taiwan’s Audrey Tang on the Limits of Biometric Identity
Key Highlights
- Audrey Tang says biometric verification cannot eliminate Sybil attacks because verified identities can still be sold or rented.
- Combining independent credentials can make fake identities more expensive to create and maintain.
- Privacy-preserving status lists can revoke compromised credentials without directly revealing their owners.
In July 2024, people lined up in the basement of a shopping mall in Singapore to have their eyes scanned by a silver orb.
The device belonged to Worldcoin, a cryptocurrency-linked identity project co-founded by OpenAI CEO Sam Altman. It was designed to distinguish real humans from bots by creating a digital credential linked to a person’s unique biometric characteristics.
But some people were not planning to keep the accounts created through the verification process. A migrant worker told The Straits Times that workers were being paid about S$45 each to hand over control of their newly created Worldcoin accounts and tokens to intermediaries.
By September, Singapore police were investigating seven people over suspected buying and selling of accounts connected to the project.
The case exposed a problem that biometric technology alone could not resolve. An iris scan can help establish that one person enrolled once, but it cannot guarantee who controls the resulting credential afterward.
For Audrey Tang, Taiwan’s cyber ambassador and former Minister of Digital Affairs, this is one of the fundamental challenges facing digital identity systems. “Nothing prevents Sybil attacks outright, biometrics included,” Tang told The Crypto Times in an exclusive email interview.
An iris scan, Tang explained, can establish that a particular body has enrolled once. However, it cannot guarantee that the same person will continue to control the resulting digital credential. The credential can still be sold or rented to somebody else.
The vulnerability matters beyond Worldcoin. It raises questions about how governments, cryptocurrency networks and decentralized organizations can distinguish genuine participants from those operating multiple identities.
As artificial intelligence makes it easier to create convincing digital personas, those questions are becoming increasingly relevant.
Why Proving Someone Is Human May Not Be Enough
A Sybil attack occurs when one person or organization creates or controls multiple identities to gain an unfair advantage in a digital system. On a blockchain, that could mean controlling thousands of wallets to influence a governance vote, manipulate an incentive program, or create the appearance of widespread community support.
In ordinary online systems, the same technique can be used to distort public consultations, manipulate ratings, or amplify particular opinions.
Biometric identification attempts to address this by associating each digital identity with a unique physical person. However, Tang argues that uniqueness is only one part of the problem. Even when the biometric scan is genuine, the person controlling the credential afterward may be somebody entirely different.
World has introduced additional protections against this possibility, including Face Auth, which uses a phone’s camera to check whether the person accessing certain features matches the individual who originally completed Orb verification.
According to World ID’s documentation, the comparison occurs on the user’s device. The feature adds a safeguard against unauthorized use, although its existence does not establish that every possible form of credential transfer or coordinated misuse has been eliminated.
Tang believes the more practical approach is to make maintaining fake identities sufficiently expensive that an attack becomes economically unattractive.
“What any system can do is raise the cost of each additional fake identity above what it is worth to the attacker,” Tang said.
Rather than relying on a single biometric scan or identity provider, decentralized systems can combine evidence from several independent sources. These might include credentials issued by different organizations, community memberships, endorsements from other participants, and records of genuine participation.
A person attempting to manufacture hundreds of identities would then need to establish convincing histories and relationships for each of them. The additional signals, however, only improve security if they are genuinely independent.
If several supposedly different credentials all depend on the same phone number or login, they may offer little additional protection against an attacker controlling that underlying account.
“Independent is the operative word,” Tang said. Forging multiple unrelated signals across many identities would increase the effort and expense involved. If coordinated abuse were discovered, the affected credentials could also be revoked.
Yet Tang does not believe every digital activity should require the same level of identity verification.
A national election, for example, demands stronger protections against duplicate participation than a public discussion about local policy. Sometimes, Tang suggested, the best protection may be to avoid unrestricted registration altogether.
This was the approach used in Taiwan during a public consultation on information integrity in March 2024.
Instead of allowing anyone to enter an open online discussion, organizers randomly sent 200,000 invitations through the government’s text-message system. According to Taiwan’s Ministry of Digital Affairs, the exercise received 1,760 valid responses. A total of 447 citizens subsequently attended the deliberative assembly, with participants selected through stratified sampling based on demographic characteristics.
The idea was to create a group broadly representative of Taiwan’s population without requiring everyone on the internet to establish a permanent digital identity.
An attacker could create additional accounts, but that would not necessarily increase the number of invitations received.
For Tang, this illustrates why the appropriate defense against fake identities may sometimes be a change in participation rules rather than a more sophisticated identity scanner.
What Happens When a Digital Identity Is Stolen?
Preventing fake identities is only one challenge. Another begins when a legitimate credential is compromised.
Imagine a person using a digital identity to access public services, prove professional qualifications, or participate in an online election. If someone steals that credential, the issuing authority must be able to invalidate it.
However, checking whether a credential has been revoked can itself create a privacy risk by revealing when and where its owner uses it.
That creates a difficult choice. A system designed to protect users against impersonation could become a system for monitoring their activities. Tang believes cryptographic mechanisms can help resolve this conflict. “Publish status, not identities,” Tang said.
Tang pointed to the World Wide Web Consortium’s Bitstring Status List, which became an official W3C Recommendation on May 15, 2025.
The technology allows credential issuers to publish revocation information without creating a publicly accessible list of people’s names or personal details. Instead, the issuer maintains a compressed collection of bits, with at least 131,072 positions.
Each credential is assigned a position. A verifier downloads the shared status list and inspects the relevant position to determine whether the credential remains valid. Because the check happens on the verifier’s side against the downloaded list, the issuer is not told which specific credential was checked. Caching that downloaded list across many verifications also avoids sending a fresh request to the issuer each time a credential is presented; reducing both the load on the issuer and the trail of lookups that could otherwise be used to track credential use.
For example, a service checking someone’s educational qualification could determine whether the credential had been revoked without necessarily informing the university that the individual was presenting it. However, the protections depend on how the system is built.
Tang warned that a credential’s position in the status list is a stable identifier. If that information is repeatedly disclosed, separate services may be able to correlate the use of the same credential.
There is also the possibility that an issuer could deliberately undermine the system. An organization could publish a separate status list for each credential, allowing it to recognize which individual’s status was being checked.
The W3C specification explicitly discusses these risks and recommends measures to reduce tracking, including shared lists, caching, and careful allocation of status entries.
Tang suggested periodically reissuing credentials with fresh positions to limit correlation.
For stronger privacy requirements, cryptographic accumulators provide another possibility. They can allow a credential holder to prove, using zero-knowledge techniques, that a credential has not been revoked without revealing the status-list position associated with it.
A simpler alternative is to issue credentials with short expiration periods. Rather than explicitly revoking a compromised credential, an institution could decline to renew it when it expires. The limitation is that the issuing institution must remain available to handle renewals.
Tang believes the rights of the person holding the credential must remain part of the design. “Whatever the mechanism, holders should be notified when a credential is revoked and given a way to appeal,” Tang said.
That requirement becomes particularly important when digital credentials determine whether individuals can access essential services or exercise legal rights.
What if the Government That Issued Your Identity Disappears?
There is another problem that receives less attention. What happens to a digital identity when the institution that issued it no longer exists?
Consider a university that closes, a government identity service that is discontinued or an organization whose servers permanently go offline.
Millions of digital credentials could still carry valid cryptographic signatures. But the systems needed to verify their status might no longer be operating.
A signature can establish that a credential was issued using a particular private key. Verification is possible using the corresponding public key. However, determining whether the credential remains valid may also require access to updated revocation information. Without it, a verifier could be unable to determine whether a credential had been compromised or canceled.
Tang believes decentralized identity systems can survive such situations, but only when that possibility is considered from the beginning. “Yes, if survival is planned for,” Tang said.
The first step is to ensure that public keys and credential status lists are not stored exclusively on the original issuer’s infrastructure.
Copies can be maintained by independent organizations, allowing verification to continue even after the issuer’s servers disappear. Credential holders can also carry copies of status lists themselves, an arrangement permitted by the W3C standard.
This avoids requiring every verifier to retrieve information directly from the original issuing authority.
Tang also highlighted a more subtle risk involving domain names. If the public keys needed to verify a credential can only be discovered through the issuer’s website, an expired domain could be acquired by somebody else. That could undermine the trust assumptions on which the identity system was built.
Independent archival arrangements and properly authenticated records can help preserve the information needed for verification.
But preserving the ability to verify old credentials does not mean the issuer’s functions can be fully replaced. A defunct institution can no longer issue new credentials, renew expired ones, or revoke credentials that become compromised after its closure.
“A frozen status list means a credential compromised after the shutdown can never be revoked,” Tang warned.
The consequence is that an apparently valid credential might remain usable even though the original authority would have invalidated it had it still been operating.
To address this, Tang proposed that institutions publish signed notices before permanently shutting down, archive their final public keys and status records, and arrange for independent parties to preserve them. The authority responsible for maintaining revocation information could also be transferred to a successor organization.
Yet such arrangements cannot entirely replace an issuer’s ability to continue verifying information or providing updated credentials.
For credentials with expiration dates, a defunct identity system may gradually lose its usefulness rather than collapse immediately.
For Tang, the longer-term solution is to avoid building digital identities around a single institution in the first place. A person might hold overlapping credentials from governments, educational institutions, professional organizations, and communities. If one issuer disappears, other credentials may still provide useful evidence of that person’s identity or qualifications.
The underlying principle is that no single organization should have complete control over an individual’s ability to prove who they are.
That brings the discussion back to the problem exposed by the Worldcoin account-trading cases.
A highly sophisticated biometric scanner cannot guarantee that the person who enrolls will remain the person who controls the resulting account. Similarly, a mathematically valid credential cannot guarantee that the institution responsible for it will remain operational forever.
Cryptography can establish authenticity, verify signatures, and protect sensitive information. But it cannot, by itself, eliminate financial incentives to cheat or ensure institutions continue to function.
The challenge for decentralized identity, as Tang’s responses suggest, is not simply to make digital credentials difficult to forge. It is to create systems that remain trustworthy when people misuse them, credentials are stolen and even the institutions that created them cease to exist.
Also Read: Security Lapses, Infra Breaks, and Regulatory Heat on Indian Crypto Exchanges: What’s Happening?
Disclaimer: The information researched and reported by The Crypto Times is for informational purposes only and is not a substitute for professional financial advice. Investing in crypto assets involves significant risk due to market volatility. Always Do Your Own Research (DYOR) and consult with a qualified Financial Advisor before making any investment decisions.


