Security researchers and cryptographers have long relied on automated tools to reason about the security of complex cryptographic protocols. Among the most popular and influential of these tools are ProVerif and Tamarin, which are widely used to analyze protocols that secure everything from messaging applications to internet infrastructure. For years, the community has referred to these systems as "protocol provers," a title that implies they can mathematically guarantee a protocol is secure. However, a new research paper argues that this terminology is not only misleading, but fundamentally dangerous.
In a newly published paper on the Cryptology ePrint Archive (Paper 2026/2683) titled "Stop Calling Them Provers: The Case for Calling ProVerif and Tamarin Symbolic Protocol Analyzers," researcher Nadim Kobeissi of Symbolic Software raises serious concerns regarding the theoretical foundations, implementation correctness, and practical limitations of these tools. The paper contends that labeling ProVerif and Tamarin as "provers" creates a false sense of absolute security, masking the inherent gaps between theoretical soundness arguments, software execution, and real-world protocol deployment.
Deconstructing the Foundations of Protocol Security
To understand the core argument of the paper, one must look closely at what is required to claim that a cryptographic protocol has been "proved" secure. According to Kobeissi, such a claim does not rely on a single factor, but rather on a delicate chain consisting of the underlying mathematical theory, the analysis software itself, the formal model of the protocol, and the specific security property being checked. If any single link in this chain fails or is misunderstood, the entire proof collapses.
To test the strength of these links, the author undertook a detailed audit of the published soundness arguments for both ProVerif and Tamarin. The findings of this audit are stark. The review uncovered false statements, missing proofs, and notable gaps between the procedures that were justified on paper and the actual operations executed by the software tools in practice. Furthermore, the paper successfully reproduced previously reported bugs that cause Tamarin to accept false claims regarding the security models submitted by users.
The software architecture of these tools also introduces trust assumptions that run counter to the idea of an infallible mathematical proof. For instance, ProVerif supplies no verifiable certificate for its full analysis, requiring users to place a high degree of trust in the tool’s internal execution. Tamarin, on the other hand, replays its proof tree using its own internal solver, which ultimately means the correctness of that validation code must also be trusted blindly. These procedural and implementation realities highlight why describing the tools as automated provers overlooks critical points of potential failure.
The Gap Between Symbolic Results and Real-World Security
Even when ProVerif or Tamarin functions perfectly and yields a correct symbolic result, the paper emphasizes that this success can still fall dangerously short of the actual security guarantees required for a protocol in the real world. Symbolic analysis operates within an abstraction layer that often fails to capture the messy reality of software implementation and deployment.
For example, protocol models can easily omit critical behaviors found in real-world parsers and state machines, leaving blind spots that attackers can exploit. Secrecy queries, which are meant to verify that sensitive data remains hidden, can occasionally be formulated in ways that still allow subtle information leakage. Additionally, the mathematical equations used to represent cryptographic primitives—such as encryption, signatures, and hashing—require rigorous justification before they can genuinely support strong computational guarantees in actual code.
While the paper acknowledges that better models, stronger queries, and computational soundness arguments can successfully address many of these problems, it notes that none of these safeguards happen automatically. Someone must actively establish that these advanced techniques actually apply to the specific protocol being analyzed. Automated tools cannot bridge this gap on their own.
A Call for More Accurate Terminology and Reporting
Because of these theoretical shortcomings, implementation bugs, and modeling limitations, the paper argues that continuing to refer to ProVerif and Tamarin as "provers" is misleading to developers, auditors, and the broader cryptographic community. Instead, the author advocates for a shift in nomenclature, suggesting they be accurately designated as "symbolic protocol analyzers."
Alongside this renaming, the paper urges researchers and engineers to change how they report their findings. Rather than publishing vague claims that a protocol has been "proven secure," results should always be reported alongside the exact models, underlying assumptions, and concrete evidence that support them.
Despite these strong criticisms and the call for reclassification, the paper does not dismiss the value of these tools. ProVerif and Tamarin remain exceptionally useful instruments for finding practical attacks and establishing symbolic properties, including the verification of security guarantees even across unbounded numbers of sessions. By recognizing them for what they truly are—sophisticated symbolic analyzers rather than infallible provers—the security community can use them more responsibly and build more resilient cryptographic systems.






