What You See Is (Not) What You Got: The Human Factor in PDF Signature Verification
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: "What You See Is (Not) What You Got".
Elias: Here's the gist of the paper. Signed PDF documents have legally binding value under frameworks such as the EU’s eIDAS regulation.
Nadia: In this work, we present Genuine PDF Forgeries, a novel attack class that focuses on human factors in signature validation.
Elias: We find that nine of the 13 viewers that automatically report signature validity hide authenticated signer information behind additional user interaction.
Nadia: Attack Formulation The authors introduce Genuine PDF Forgeries, an attack class that exploits the gap between cryptographically validated signer information and human-visible signature blocks The attacks present forged signer information in the visual signature block while attaching a cryptographically valid PDF signature.
Title and authors: Nadia: So we're looking at this paper now, "What You See Is (Not) What You Got: The Human Factor in PDF Signature Verification," and it seems like the core issue here is a disconnect between what's cryptographically true and what the user actually sees.
Elias: Yeah, that’s right. The authors are focusing on how human factors mess up signature validation when you have legally binding documents under rules like eIDAS. They argue that signed PDFs have legal weight, but users don't always know if they're looking at the real signer or something else.
Priya: What I’m curious about is what kind of stuff they’re focusing on. Is this just about digital signatures in contracts, or does it apply to things where people are making high-stakes decisions?
Nadia: It applies broadly, Priya. They introduce a new attack class called Genuine PDF Forgeries. The idea is to put fake signer information right into the visual signature block while keeping the underlying PDF signature actually valid cryptographically.
Elias: And they systematize this by creating three main variants: forged badges attached to a real signature, forged elements that just sit next to a valid one, and even forging elements in documents that aren't signed at all.
Priya: So, if a document looks right on the surface—valid signature present—but the person looking at it is being tricked by what they see versus what's actually underneath, how much risk are we talking about?
Nadia: That’s exactly the point. They surveyed twenty-three different PDF viewers and validation tools to see how they handle these forged documents in real-world software. The finding was that nine out of those thirteen viewers that automatically give you a yes or no answer about the signature being valid actually hide the real signer's certificate details until you manually dig for them.
Elias: It’s a very specific human factor problem, Elias. They found that this kind of user experience is misleading most of the time when these tools are subjected to their attacks.
Title and authors: Priya: That sounds like it could be dangerous if someone relies on that immediate visual confirmation without actually checking the certificate details underneath. How much does this change how we trust a document?
Nadia: Well, the researchers ran deception-based online studies with five hundred eighty-four people who were eligible for eIDAS signatures. They found that in the most effective attack variant, seventy one point nine percent of those participants were tricked on their first impression by the forged information.
Elias: And even after they got prompted to look deeper, fifty seven point three percent of them were still fooled when they looked further into the document. It shows that simply showing a boolean validation result isn't enough to guarantee trust in this context.
Priya: So what are the suggested fixes for this problem? The authors propose two main ways to fight it off, and I want to know what those look like in practice.
Nadia: They suggest attacking it from two angles. First, viewers should show information that actually contradicts the forged signature block by overlaying it with clear indications of validity and trustworthy signer details from the certificate itself.
Elias: And the second approach is about reducing how much users rely on just looking at that signature block in general, which is a big design change. They prototype a way to overlay the untrusted visual signature block with the real signer information pulled directly from that validated certificate.
Priya: That overlay idea sounds like it’s trying to make sure the system presents trustworthy information immediately rather than hiding it behind extra steps for the user. But if you hide it, doesn't that just shift the trust problem somewhere else?
Nadia: It shifts it, maybe, but they argue that making authenticated signer attribution part of the default validation interface instead of hiding it behind optional user interaction is the final recommendation. That means seeing who signed a document should be front and center by default.
Elias: That ties back to their attack formulation where they specifically target that gap between what's cryptographically verified and what's human-visible. They’re saying we need to close that gap by making the real information impossible to ignore.
Title and authors: Priya: So, if we look at this from a privacy standpoint, if the viewer is forced to show the real certificate details right away, are we risking more exposure or less?
Nadia: The paper focuses on making sure that what you see is trustworthy information rather than something an attacker controls. The goal is to make users better equipped to spot the forgery when they are looking at a document with potentially forged visual elements.
Elias: It sounds like the research points toward a design solution where authenticity isn't optional; it should be part of the default validation screen for any signed PDF.
Priya: And for me, what this means is that we need to think about how these verification interfaces are designed from the start to prioritize user trust through transparent information display. That makes sense given how easily people can be tricked by visual cues.
Nadia: Exactly. So, to wrap up on "What You See Is (Not) What You Got: The Human Factor in PDF Signature Verification," the paper shows that the combination of a boolean validity result and an attacker-controlled visual signature block creates a high risk of user confusion, especially in sensitive areas.
Elias: They demonstrated that prominent boolean messages can actually reinforce the forgery instead of exposing it, which is a key finding for anyone building these verification tools. The core idea is to move signer attribution out of the optional interaction step.
Priya: It really highlights that the trust mechanism isn't just in the cryptography; it’s in how that cryptography is presented to the end-user.
Nadia: That’s what this paper suggests, Priya, and it points us toward designing interfaces where authenticated signer attribution is part of the default validation interface. That should give users a much stronger footing when they are looking at those legally binding documents.
Elias: For anyone building verification systems, the message is clear: don't hide the real identity behind extra steps if you want to avoid these kinds of human-factor exploits. That’s what Genuine PDF Forgeries shows us.
The paper's summary: Nadia: So, this paper is about how people actually see PDF signatures versus what’s secretly happening underneath.
Elias: Right, it focuses on this gap between what a computer knows is true and what a human user perceives when checking a legally binding document.
Priya: The authors found that attackers can exploit this by putting fake signer details right into the visual signature area while keeping the actual signature cryptographically valid.
Nadia: Exactly. They call it Genuine PDF Forgeries, and they made three specific types of attacks, like forged badges or just elements that sit next to a real signature.
Elias: And then they surveyed twenty-three different PDF viewers and validation tools to see how they handle this in the real world.
Priya: The main finding there is that nine out of those thirteen viewers that give you an automatic yes or no about the signature being valid, they actually hide the real signer information until you manually click into it.
Nadia: That’s a big deal for user trust because if someone just sees a green checkmark, they might think it's legitimate without ever checking the certificate details.
Elias: And in their deception studies with five hundred eighty-four people, the most effective way to trick them was that initial visual impression.
Priya: They found that seventy one point nine percent of participants were fooled on first look by this forged information, and even after asking them to look closer, fifty seven point three percent were still tricked.
Nadia: So the implication for me is that just showing a boolean result isn't enough to make people trust these documents when they’re in high-stakes situations.
Elias: And the authors propose countermeasures, like overlaying the untrusted signature block with real signer info from the certificate, or making sure that real attribution is part of the default validation screen.
Priya: So it's not just about cryptography anymore; it's about designing interfaces where trustworthy information is presented immediately instead of hiding it behind extra steps for the user.
Nadia: That makes sense. The core issue they highlight is that prominent boolean messages can actually reinforce a forgery instead of exposing it.
Elias: It points toward a design principle where the real identity should be front and center by default, not optional or hidden away.
Priya: And that’s what we need to think about when designing these verification tools—making sure the user sees what’s trustworthy right away.
The paper's improvements: Tom: So, this part is about how the authors suggest fixing these problems they found in viewers and validation tools.
Nadia: They suggest two main ways to fight this attack: one is making viewers show information that actually contradicts a forged signature block by overlaying it with real details from the certificate.
Elias: That’s trying to resolve the visual confusion by presenting trustworthy information directly, instead of hiding it behind extra steps for the user.
Priya: I see that as trying to make users better equipped to spot a forgery when they are looking at a document with potentially fake elements in the signature.
Nadia: And they also suggest reducing how much users rely on just looking at that signature block in general, which is a big design change for interfaces.
Elias: They prototype this by overlaying the untrusted visual signature block with real signer information pulled directly from that validated certificate so it’s always visible.
Priya: So they are pushing for authenticated signer attribution to be part of the default validation interface, not something you have to hunt for manually.
Nadia: That seems like a big shift because it means making the real identity impossible to ignore by putting it right in front of the user.
Elias: It ties back into their main point: that authenticated signer attribution should be part of the default validation interface instead of hidden behind optional user interaction.
Priya: From a privacy standpoint, if you force that display, are you risking more exposure or less when dealing with those legally binding documents?
Nadia: The goal is to make sure what the user sees is trustworthy information rather than something an attacker controls by making the real certificate details prominent.
Elias: They argue this approach can actually make users better equipped to detect the attack because it closes that gap between what’s cryptographically verified and what’s human-visible.
Priya: So it really boils down to designing systems where authenticity isn't optional; it should be part of the default validation screen for any signed PDF.
Conclusion: Nadia: So we're wrapping up on "What You See Is (Not) What You Got: The Human Factor in PDF Signature Verification."
Elias: It’s basically about how human interaction messes with signature validation when you have legally binding documents under rules like eIDAS.
Priya: The core finding is that the combination of a boolean validity result and an attacker-controlled visual signature block creates a high risk of user confusion, especially in sensitive areas.
Nadia: Yeah, they showed that nine out of thirteen viewers hide the real signer identity until you do extra work, and their deception studies showed seventy one point nine percent were tricked on first impression.
Elias: It highlights that we need to think about how these verification interfaces are designed from the start to prioritize user trust through transparent information display.
Priya: I think it really points toward designing systems where authenticated signer attribution is part of the default validation interface instead of hiding it behind optional interaction.
Nadia: That’s the final recommendation: make sure seeing who signed something is part of the standard screen, not an extra step for the user.
Elias: It’s about closing that gap between what a computer knows is true and what a human user actually sees when checking those documents.
Priya: So, it shifts the trust mechanism from just being in the cryptography to how that cryptography is presented to the end-user.
Jakob Heher, Emanuel Pichler, Stefan More
Institute of Information Security, Graz University of Technology · Secure Information Technology Center Austria (A-SIT)
cs.CR, cs.HC
Submitted: 2026-11-11
Updated: 2026-11-11
DOI: 10.1016/j.cose.2026.105191
Code: https://github.com/pwmt/zathura-pdf-poppler
License: http://creativecommons.org/licenses/by-nc-nd/4.0/
Importance score: 67/100
The gist: Signed PDF documents have legally binding value under frameworks such as the EU’s eIDAS regulation.
Key concepts
- Genuine PDF Forgeries
- This attack class involves placing fake signer information in a visible signature block while attaching a technically valid PDF signature. It targets the gap between verified digital data and what the user actually sees.
- Viewer Survey Findings
- Testing 23 PDF viewers revealed that nine of them hide authenticated signer certificate details unless the user manually navigates to them after an automatic validity check. This indicates a misleading user experience for many validation tools.
- Deception-Based User Studies
- User studies demonstrated that attackers can deceive nearly 72% of participants on first impression, and this deception often persists even after users are prompted to look closer. This highlights the vulnerability in real-world use scenarios.
Terminology
Summary
Signed PDF documents have legally binding value under frameworks such as the EU’s eIDAS regulation. p. 0 What You See Is (Not) What You Got:
The Human Factor in PDF Signature Verification
Jakob Heher a,b,∗
, Emanuel Pichler a
, Stefan More a,b
a
Institute of Information Security, Graz University of Technology, Graz, Austria
bSecure Information Technology Center Austria (A-SIT), Graz, Austria
Abstract
Signed PDF documents have legally binding value under frameworks such as
the EU’s eIDAS regulation. In this work, we present Genuine PDF Forgeries,
a novel attack class that focuses on human factors in signature validation. What You See Is (Not) What You Got:
The Human Factor in PDF Signature Verification
Jakob Heher a,b,∗
, Emanuel Pichler a
, Stefan More a,b
a
Institute of Information Security, Graz University of Technology, Graz, Austria
bSecure Information Technology Center Austria (A-SIT), Graz, Austria
Abstract
Signed PDF documents have legally binding value under frameworks such as
the EU’s eIDAS regulation. In this work, we present Genuine PDF Forgeries,
a novel attack class that focuses on human factors in signature validation. We analyze our attacks against 23 PDF viewers and validation tools across
desktop platforms and the web. We find that nine of the 13 viewers that automatically report signature validity hide authenticated signer information behind
additional user interaction.Where the paper says it
Where the paper says it
Where the paper says it
Attack Formulation
The authors introduce Genuine PDF Forgeries, an attack class that exploits the gap between cryptographically validated signer information and human-visible signature blocks The attacks present forged signer information in the visual signature block while attaching a cryptographically valid PDF signature. They systematize the class and derive three variants: forged badges that are connected to a valid signature, forged signature-like elements that coexist with a valid signature, and forged signature-like elements in unsigned documents
Viewer Survey
The authors evaluate their assumption against real-world viewers by surveying 23 PDF viewer applications and validation tools. They find that most viewers offering eIDAS signature validation have misleading UX when subjected to their attack: they appear to confirm the forged signature block as “valid”, while the genuine, verified signer identity remains hidden. p. 2 We check if they validate
PDF signatures, if they support eIDAS qualified trust roots, how they display
validation results in general, and verified signer information in particular. We find that most viewers offering eIDAS signature validation have misleading
UX when subjected to our attack: they appear to confirm the forged signature
block as “valid”, while the genuine, verified signer identity remains hidden.Where the paper says it
User Studies
The research conducts three deception-based user studies: one high-specificity, conducted under real-world conditions in highly privacy-sensitive areas; two high-yield, measuring the impact of the attack across a wide variety of victims. In the online studies (total n=584), the most effective attack variant deceives 71. p. 0 Five of six organizational trials accepted requests authenticated using the forged visual identity. In the online studies (total
n=584), the most effective attack variant deceives 71. In the online studies (total
n=584), the most effective attack variant deceives 71. 9% of eligible participants
on first impression, and 57. 9% of eligible participants
on first impression, and 57. 3% continue to be deceived even after we prompt for
additional introspection.Where the paper says it
Where the paper says it
Where the paper says it
Countermeasures
The authors propose mitigations attacking the problem from two fronts: one, viewers prominently providing information to contradict a forged signature block; and two, reducing users’ reliance on the signature block in general They prototype a safeguard that overlays the untrusted signature block with signer information obtained from the validated certificate. The final recommendation is that authenticated signer attribution should be part of the default validation interface rather than hidden behind optional user interaction. p. 0 We analyze our attacks against 23 PDF viewers and validation tools across
desktop platforms and the web. We find that nine of the 13 viewers that automatically report signature validity hide authenticated signer information behind
additional user interaction. We analyze our attacks against 23 PDF viewers and validation tools across
desktop platforms and the web. We find that nine of the 13 viewers that automatically report signature validity hide authenticated signer information behind
additional user interaction.Where the paper says it
Where the paper says it
Discussion
The high-specificity user study highlights that validators in highly-sensitive areas are successfully misled by the combination of a boolean validity result and an attacker-controlled visual signature block. The authors note that eIDAS-compliant viewers that immediately show a boolean result but hide signer information behind manual interaction offer the highest risk of user confusion. p. 0 We analyze our attacks against 23 PDF viewers and validation tools across
desktop platforms and the web. We find that nine of the 13 viewers that automatically report signature validity hide authenticated signer information behind
additional user interaction. We analyze our attacks against 23 PDF viewers and validation tools across
desktop platforms and the web. We find that nine of the 13 viewers that automatically report signature validity hide authenticated signer information behind
additional user interaction.Where the paper says it
Where the paper says it
Mitigations
The first approach is to resolve the visual confusion by presenting trustworthy information to the user by overlaying it with a clear indication of the signature’s validity, decorated with trustworthy signer information from the signer certificate. The authors argue that this approach can make users better equipped to detect their attack. p. 2 They attack the problem from two fronts:
one, viewers prominently providing information to contradict a forged signature
block; and, two, reducing users’ reliance on the signature block in general. We
argue that these approaches can make users better equipped to detect our attack. What You See Is (Not) What You Got:
The Human Factor in PDF Signature Verification
Jakob Heher a,b,∗
, Emanuel Pichler a
, Stefan More a,b
a
Institute of Information Security, Graz University of Technology, Graz, Austria
bSecure Information Technology Center Austria (A-SIT), Graz, Austria
Abstract
Signed PDF documents have legally binding value under frameworks such as
the EU’s eIDAS regulation. In this work, we present Genuine PDF Forgeries,
a novel attack class that focuses on human factors in signature validation.Where the paper says it
Where the paper says it
Conclusions
In this work, they introduce Genuine PDF Forgeries, an attack class that exploits the separation between certificate-derived signer identities and document-controlled visual signature information Across 23 PDF viewers and validation tools, they find that nine of the 13 viewers that automatically display a validity result conceal the authenticated signer identity until the user performs additional interaction. They demonstrate that prominent boolean validity messages can reinforce, rather than expose, a forged signer identity. p. 0 We derive viewer-side countermeasures and prototype a safeguard that overlays the untrusted signature block with signer information obtained from the
validated certificate. Our results show that boolean validity indicators, not
associated with the authenticated signer identity, are insufficient in practice.Where the paper says it
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
-
Prominent display of authenticated signer identity in validation interfaces can prevent forgery success by presenting
trustworthy information to the user
instead of hiding it behind manual interaction, which is cited as a way to mitigate attack variant A1. -
AI systems can be improved by implementing a safeguard that
overlays the untrusted signature block with signer information obtained from the validated certificate,
directly addressing the core mismatch exploited by Genuine PDF Forgeries. -
Validation tools and PDF viewers should be modified to prominently display signer identities derived from validated certificates, ensuring
authenticated signer attribution should be part of the default validation interface rather than hidden behind optional user interaction.
-
AI systems can develop mechanisms to distinguish between certification signatures and approval signatures, adopting the behavior where
certification signatures prominently display the signer’s name and the certificate issuer in the signature status bar
for approval documents. -
Systems should be designed to clearly distinguish trusted information from document-controlled content, making it more apparent to users when they are presented with a forged visual signature block that claims a
forged
signer identity different from the actual signer.
Abstract
Signed PDF documents have legally binding value under frameworks such as the EU's eIDAS regulation. In this work, we present Genuine PDF Forgeries, a novel attack class that focuses on human factors in signature validation. Our attacks target the disconnect between cryptographically verifiable signer certificates and human-visible signature blocks embedded in documents. We analyze our attacks against 23 PDF viewers and validation tools across desktop platforms and the web. We find that nine of the 13 viewers that automatically report signature validity hide authenticated signer information behind additional user interaction. We then evaluate the attack in twin deception-based online studies with 584 EU-based participants. To further validate the impact, we perform a highly tailored real-world attack study with six trials within two privacy-sensitive organizations. Five of six organizational trials accepted requests authenticated using the forged visual identity. In the online studies (total n=584), the most effective attack variant deceives 71.9% of eligible participants on first impression, and 57.3% continue to be deceived even after we prompt for additional introspection. We derive viewer-side countermeasures and prototype a safeguard that overlays the untrusted signature block with signer information obtained from the validated certificate. Our results show that boolean validity indicators, not associated with the authenticated signer identity, are insufficient in practice.
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