Clear Human Web: Where a Person Vouches for Every Client
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Security Radio. Generated commentary on the latest security and cryptography papers.
Nadia: We're looking at the paper: "Clear Human Web: Where a Person Vouches for Every Client".
Elias: Here's the gist of the paper. The paper proposes the Clear Human Web, a part of the public web where sites open only to clients a person vouches for.
Nadia: This construction aims to reveal whether a person stands behind the client, which reveals nothing else.
Elias: The Problem Addressed The problem has three faces, one for each party on the web.
Nadia: Whoever publishes has no fence.
Paper summary: Nadia: We're now getting into how this paper actually sets up that web. They start by pointing out that on the existing web, publishers have no fence, measurers have no count, and readers have no proof of identity <ref:2611.06286#pg2>.
Elias: Exactly. The fundamental issue is a web that cannot see the person who is making the request. They call for one cure: a check that a person stands behind the client, which reveals nothing else about them <ref:2611.06286#pg3>.
Priya: So, what's the proposed mechanism again, focusing only on what's in the summary? I want to make sure we get the high-level idea right before we get into how it runs.
Nadia: The construction involves a person’s wallet presenting their personal identification data to an issuer. This issuer is the only party that acts as a relying party and hands back a bounded stock of anonymous proofs of personhood. Then, the person's browser spends one proof at every site that runs the gate when those sites ask for it <ref:2611.06286#pg1>.
Elias: The site only checks that the proof came from an issuer it trusts and opens its door, but it doesn't learn who the person is, and it can't tell if two visits come from the same person <ref:2611.06286#pg1>. Nothing changes in the protocols of the web or in what publishers are obligated to do.
Priya: That sounds like a really clean separation of concerns. It suggests that we don't need to overhaul every single protocol just to get this one specific verification layer working properly <ref:2611.06286#pg2>.
Nadia: Right, and the paper lays out five properties they have to satisfy: admission by persons, no identity for the person, unlinkability, bounded multiplicity for visits, and delegation tolerance <ref:2611.06286#pg3>. They show that an optimum solution exists where k equals one token per site and no tokens at all <ref:2611.06286#pg3>.
Elias: And they prove this optimum works against seven different adversaries, which shows the construction is quite robust under various attack models <ref:2611.06286#pg3>. This gives us a solid benchmark for what's achievable in this space.
Priya: So, before we move on to the conclusion and title implications, what’s the main takeaway from all this setup? What is the practical meaning of "Clear Human Web"?
Nadia: The main point is that one bit of information—that a person stands behind a client—is enough to let a part of the public web emerge wherever sites choose to join it <ref:2611.06286#pg1>. It’s about making that verification happen without needing a new root of trust from scratch.
Elias: It’s essentially using the EU wallet to supply that bit, which is a blind RSA signature over a random value that carries no attribute <ref:2611.06286#pg3>.
Conclusion: Nadia: So we’re wrapping up our discussion on "Clear Human Web: Where a Person Vouches for Every Client." The authors are Lorenzo Porcelli and his independent research group <ref:2611.06286#pg1>.
Elias: And the title itself suggests that the core value isn't just verification, but the ability for sites to join this system based on a single vouching person <ref:2611.06286#pg3>. It’s about defining a specific structure where trust is delegated by an individual, not by some centralized authority.
Priya: From my perspective as someone focused on privacy, what does this mean for the average user who just wants to browse the web without their identity being tracked everywhere?
Nadia: It means that instead of every site trying to figure out if you're a human or an automated script, they can rely on this specific mechanism where you only need to prove you have a valid token from an issuer <ref:2611.06286#pg3>. The system only makes the enforcement happen per person rather than per address <ref:2611.06286#pg3>.
Elias: That distinction is important because the optimum construction they found removes that linking problem when you use Tor for example, where the address itself says nothing about who is using it <ref:2611.06286#pg3>. The gap between a site that can't report and an issuer that can't see is what this per-site pseudonym closes <ref:2611.06286#pg3>.
Priya: And the limitations they state are pretty clear too, right? What’s the practical stop for this approach?
Nadia: They point out a few limits. First, you still need someone with a PID in their wallet to participate <ref:2611.06286#pg2>. Second, the construction today inherits links from the web even without Tor <ref:2611.06286#pg3>. And finally, they mention that two extensions follow from the same proof, like writing or voting, which needs a visitor that a site can recognize and refuse without naming them <ref:2611.06286#pg3>.
Elias: So the next step toward that optimum involves adding A zero tokens in the browser to handle those extensions <ref:2611.06286#pg3>. It’s an iterative process, not a single fix for everything.
Priya: It sounds like a very focused piece of research that defines a specific, verifiable layer on top of the existing infrastructure rather than trying to replace everything at once.
Nadia: That’s right. We saw how this paper defines the Clear Human Web through five properties and seven adversaries and showed which ones the construction they proposed satisfies against each adversary <ref:2611.06286#pg3>. It shows exactly where this concept fits in terms of what it can do for the web as a whole, without overpromising on what it can achieve immediately <ref:2611.06286#pg3>.
Lorenzo Porcelli
cs.CR, cs.CY, cs.NI
Submitted: 2026-11-04
Updated: 2026-11-04
Code: https://github.com/eu-digital-identity-wallet/eudi-doc-attestation-rulebooks-catalog
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 83/100
The gist: The paper proposes the Clear Human Web, a part of the public web where sites open only to clients a person vouches for.
Key concepts
- Clear Human Web
- A part of the public web where access is restricted based on a person vouching for it. This system reveals only whether a person stands behind a client, offering accountability without exposing other identifying information.
- Anonymous Proofs
- Digital tokens issued by an issuer that a person holds in their wallet. A site checks these proofs to ensure they come from a trusted source but does not learn the user's identity or track multiple visits from the same person.
- P1 Admission by persons
- A property requiring that sites can be admitted into the system by individuals. The construction aims to satisfy this, ensuring that people can use the web without needing a new root of trust for every site they visit.
Summary
The paper proposes the Clear Human Web, a part of the public web where sites open only to clients a person vouches for. p. 0 This construction aims to reveal whether a person stands behind the client, which reveals nothing else. p. 1
The Problem Addressed
The problem has three faces, one for each party on the web. Whoever publishes has no fence. p. 0 Whoever measures has no count. p. 0 Whoever reads has no proof. p. 0 The core issue is a web that cannot see the person, and call for one cure, a check that a person stands behind the client, which reveals nothing else. p. 1
Construction Steps
The construction has three steps. A person’s wallet shows the PID to an issuer, which verifies it and hands back a bounded stock of anonymous proofs. p. 1 The person’s browser spends one proof at each site it visits, when the site asks for it. p. 1 The site checks that the proof comes from an issuer it trusts and opens the door; it does not learn who the person is, and cannot tell that two visits come from the same person. p. 1 Nothing changes in the PID, in the protocols of the web or in the obligations of the sites. p. 1
Properties and Objectives
The construction must satisfy five properties. These properties are P1 Admission by persons, P2 No identity, P3 Unlinkability, P4 Bounded multiplicity, and P5 Delegation tolerance. p. 3 The optimum satisfies these properties with k = 1 and no tokens at all. p. 3 For instance, a person counts once on every site, and there is no census. p. 3
Analysis of Performance
The analysis shows that the construction meets these objectives against seven adversaries. Table 3 collects the outcome for the construction of today and for the optimum. p. 8 For example, an honest-but-curious site satisfies P4 per pseudonym, since the wallet does not limit the pseudonyms a person holds at an issuer. p. 8 The optimum satisfies P4 with k = 1 and no issuer. p. 3
Limitations and Future Work
Three limits hold at every step, the optimum included. These include excluding anyone without a PID in a wallet, and the construction of today inherits links from the web without Tor. p. 5 The optimum removes this with Tor, where the address says nothing. p. 1 Two extensions follow from the same proof. p. 1 These include writing, as in comments, edits or votes, needs a visitor that a site can recognise and refuse without naming them, which is the per-site pseudonym of the optimum. p. 2 The next step toward the optimum requires A0 Tokens in the browser. p. 9
Conclusion
One bit of information, that a person stands behind a client, is enough to let a part of the public web emerge wherever sites choose to join it. The EU wallet supplies that bit without a new root of trust. p. 5 We have defined the Clear Human Web through five properties and seven adversaries, and shown which of them the construction of today satisfies. p. 0 The publisher’s agent asks an issuer to blind-sign the site’s domain name under the issuer’s badge key of the epoch. p. 7 The token is a publicly verifiable type of Privacy Pass, a blind RSA signature over a random value, and carries no attribute. p. 6 The epoch is the lifetime of an issuer key. p. 6 Every site counts her as one visitor: with k = 1 she cannot be two visitors on A. p. 3 The bound holds per site, and a person pays one PID presentation for every further k. p. 5 The gap between the two, a site that cannot report and an issuer that cannot see, is what the per-site pseudonym of the optimum closes. p. 0 The construction only makes them enforceable per person rather than per address. p. 5 The optimum removes it with Tor, and the observer learns only that someone uses Tor. p. 1
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Attestation of personhood through zero-knowledge proofs allows a site to admit persons while ensuring
No identity
is learned by any party, asThe issuer learns only that some PID holder obtained proofs.
-
The system can enforce a strict visitor bound on any single site and epoch, stating that
On any one site and within one epoch, the validity period of a proof, a person can appear as at most k distinct visitors
with k=1 in the optimum. -
A publisher of a site gains protection because
The publisher of each site holds the same proof as its readers,
creating an enforced fence thatholds by cryptography where the Robots Exclusion Protocol only asks.
-
The system can distinguish between a person and an agent by bounding delegated access, as
A program has the identities of the person behind it, and no more,
ensuring a program is admitted becausea person vouches for it.
-
The construction can be optimized to have no issuer by using pseudonyms derived from the proof itself, allowing
a site can refuse a pseudonym for the epoch without learning who it is
against adversaries like the network observer.
Sources
- Breaking reCAPTCHAv2
- Broken Gates: Re-evaluating Web Bot Defenses in the Age of LLM Agents
- HLL: Can Agents Cross Humanity's Last Line of Verification?
- CAPTCHAs in the Agentic Era: Solvers That Learn from Every Encounter
- The Impact of AI-Generated Text on the Internet
- Large Language Models Pass the Turing Test
- Personhood credentials: Artificial intelligence and the value of privacy-preserving tools to distinguish who is real online
- Authenticated Delegation and Authorized AI Agents
- A Brief Note on Cryptographic Pseudonyms for Anonymous Credentials
- Security Risks of AI Agents Hiring Humans: An Empirical Marketplace Study
Related papers
- From Strategy Search to Policy Learning: MCTS-Guided Adversarial Co-Evolution for LLM Safety Alignment
- Privacy Leakage from Recurrent State in LLM Inference
- Incrementally Verifiable Oracle Aided Computation
- Privilege and confidentiality in generative AI workflows
- BadMoE: Backdooring Mixture-of-Experts LLMs via Routing-Aware Triggers and Task-Underutilized Experts
- CoVault: Secure, Scalable Analytics of Personal Data