Skip to content
INTERNET INFRASTRUCTURE & NETWORKS

Engineering the Next Generation: How the Web Is Preparing Its Post-Quantum Privacy Infrastructure

As the technology industry looks ahead to the eventual arrival of cryptographically relevant quantum computers, the primary focus for security engineers has long been concentrated on the foundational building blocks of authenticated key-agreement protocols, such as TLS. Without these core protections firmly established, secure end-to-end communication across the public internet will cease to function once quantum systems reach maturity. However, modern internet infrastructure relies heavily on classical, quantum-vulnerable cryptography for far more than just standard transport layer security. Protecting applications from abuse, mitigating volumetric attacks, and preserving user anonymity all depend on protocols that are destined to be broken by future quantum hardware unless significant engineering transformations take place.

Among the most prominent tools used to protect modern web applications without sacrificing user privacy is Privacy Pass, a protocol deployed widely by major network operators like Cloudflare. The primary function of Privacy Pass is to mitigate application abuse by implementing rate-limiting mechanisms for individual clients, ensuring that no single automated actor or misbehaving application can consume a disproportionate share of a site’s resources. Historically, solving this resource management problem involved introducing significant friction or invasive privacy trade-offs for everyday human users. Web operators frequently deployed intrusive CAPTCHAs, persistent user accounts, or complex device fingerprinting techniques to monitor and moderate traffic.

Privacy Pass was explicitly engineered to bypass this false dichotomy between friction and user privacy. The fundamental architecture relies on issuing a finite number of cryptographically blinded tokens to a client, which can later be redeemed seamlessly in exchange for uninterrupted access to a target site. Crucially, each individual token is entirely unlinkable to the specific client it was originally issued to. This structural guarantee ensures that everyday visitors can continue using web applications just as they always have, without surrendering their personal privacy or browsing history to centralized observers.

Following a decade of successful large-scale deployment—utilizing Privacy Pass frameworks for popular privacy proxies like Apple’s iCloud Private Relay and specialized infrastructure like Privacy Gateway—developers across the security ecosystem are collaborating to engineer the next generation of this critical web technology. Modern deployments face evolving challenges, notably the need to develop more sophisticated management tools for automated agentic traffic that threatens to degrade performance for human visitors, alongside efforts to foster a more open ecosystem that lowers the barrier to entry for diverse client types. Achieving these ambitious objectives requires developing powerful new capabilities that extend well beyond the current boundaries of standard Privacy Pass architectures.

Despite the looming horizon of quantum computing, security researchers note that the immediate transition outlook for anonymous credentials is surprisingly favorable. Current iterations of Privacy Pass continue to rely on classical cryptographic primitives, including RSA and elliptic curve math. In the face of a quantum adversary, an attacker would theoretically possess the capability to forge tokens at will, bypassing vital site operator defenses such as rate-limiting mechanisms. On the other hand, Privacy Pass maintains an unconditionally private design, meaning that no observer—quantum or classical—can link the redemption of a token back to its original issuance. Furthermore, the advanced authentication capabilities currently under development are being instantiated with architectural guarantees that ensure user interactions remain strictly private even after the arrival of post-quantum cryptanalysis milestones.

This current cryptographic reality provides a temporary window of breathing room, but the industry will require robust, production-ready alternatives well before quantum threats materialize. While this remains an active and rapidly moving area of academic and industrial research, the primary engineering hurdle mirrors the broader post-quantum transition: the sheer physical size of post-quantum cryptographic primitives. For example, modern post-quantum digital signatures such as ML-DSA-44 require a massive 2,420 bytes, whereas traditional ECDSA-P256 signatures consume a mere 64 bytes. Given these constraints, security architects harbor few illusions about finding seamless, drop-in cryptographic replacements in the near future. Just as the internet is adapting to post-quantum TLS requirements, engineers will have to accept primitives with significantly higher operational costs and re-engineer their underlying systems to bear that weight.

What We Need to Build: A Brief Look at MOLE

To understand the trajectory of post-quantum privacy engineering, researchers are exploring novel frameworks made possible by recent breakthroughs in blind signature design. One prominent initiative in this space is MOLE, an acronym standing for "Moderation Of unLinkable Endorsements."

At its technical core, Privacy Pass relies heavily on a blind signature scheme. This mathematical primitive allows a client to obtain a valid digital signature under an issuer’s secret key without ever revealing the underlying message content to the issuer itself. The client can subsequently reveal the message and its accompanying blind signature at a later time without establishing any traceable link back to the original issuance transaction.

In traditional Privacy Pass implementations, a valid token consists of a message paired with a blindly issued signature. To verify the authenticity of the token, a verifying server must inspect the issuer’s public key, inherently learning the specific identity of the issuing entity. While this structure effectively protects the privacy of the individual client, it does not conceal the identity of the issuer. For many practical applications, this transparency is entirely intentional. The identity of the issuer provides valuable signaling data regarding the type of client connecting to an edge network, allowing operators to make informed routing and authorization decisions. For instance, a client might present a token issued by a specific enterprise partner to cryptographically prove authorization to use a specialized virtual private network product.

However, for a vast array of broader web use cases, revealing the issuer’s identity exposes more information than is genuinely necessary to grant a client access to a resource. Under normal operating conditions, a verifier only needs mathematical proof that a token was successfully generated by some entity within a trusted ecosystem. MOLE pushes this privacy model a step further by ensuring that tokens actively hide the identity of the issuer altogether. By stripping away issuer visibility, verifiers are structurally prevented from treating incoming requests differently based on who originally endorsed the client, thereby guaranteeing equal treatment and making it significantly easier for diverse client architectures to participate.

This issuer-hiding property is achieved through a specialized cryptographic object known as an endorsement. An endorsement is initially issued by an entity called an anchor and is subsequently redeemed by a designated moderator. Prior to the redemption phase, the client and the moderator mutually agree upon an anchor set, which consists of the public keys belonging to all anchors trusted by that specific moderator. During the redemption process, the client provides a cryptographic proof demonstrating that it received an endorsement from at least one valid anchor within the designated anchor set, without ever disclosing which specific anchor performed the issuance. Upon successful verification, the moderator issues the client a MOLE credential, which is then utilized to moderate the client’s subsequent interactions with the target website.

While endorsements are specifically engineered to remain anchor-hiding, MOLE credentials themselves are tailored for specific moderators and do not require public verifiability or anchor concealment. This flexible architecture allows a wide variety of underlying cryptographic primitives—ranging from advanced blind signatures to anonymous credit tokens and rate-limited credentials—to successfully instantiate MOLE credentials.

A Feast of Cryptography and the VOLE-in-the-Head Paradigm

When designing complex cryptographic protocols, systems architects constantly navigate a delicate tension between adaptability and raw computational efficiency. Highly tailored cryptographic constructions often excel at solving a single, highly specific problem, but their rigid design makes them extraordinarily difficult to adapt when protocol requirements inevitably shift over time.

Adaptability is an absolute prerequisite for an architecture like MOLE. The framework is not designed to support just a single credential system, but rather an expansive ecosystem of interchangeable protocols. Because web traffic patterns, threat vectors, and application requirements change continuously, security engineers require modular building blocks that can be composed and recombined in diverse ways.

The post-quantum future of web privacy

Most modern post-quantum cryptographic schemes originate from broad families of mathematical constructions, yet they are meticulously tuned for narrow applications. ML-DSA serves as a prime example: the underlying cryptographic framework can theoretically support a wide variety of functions, but to achieve a compact, practical digital signature scheme, its specific operational parameters were narrowly optimized. Consequently, those exact parameters cannot guarantee security if applied to alternative cryptographic use cases.

Overcoming this constraint has become the central challenge of post-quantum anonymous credentials: successfully decoupling parameter selection from broader protocol design. Classical credential systems historically took this luxury for granted, allowing protocol engineers to build complex applications without needing deep, specialized expertise in the underlying mechanics of elliptic curves. Unfortunately, engineering efficient post-quantum anonymous credentials has traditionally demanded an extraordinary level of specialized mathematical expertise.

That landscape shifted significantly following a series of major cryptographic breakthroughs centered around a zero-knowledge proof framework known as VOLE-in-the-Head, originally introduced by academic researchers at Crypto 2023. A zero-knowledge proof system enables a prover to demonstrate mathematical knowledge of a hidden witness that satisfies a public statement, while ensuring that the underlying witness data remains entirely private. For example, a zero-knowledge proof can be deployed to demonstrate knowledge of a secret encryption key without revealing the key itself. When transformed appropriately, these proofs form the foundation of digital signature schemes like FAEST, which has emerged as a prominent candidate in NIST’s ongoing post-quantum signature standardization process.

Compared to many contemporary, highly complex proof systems, the VOLE-in-the-Head approach is relatively straightforward to understand and implement securely. The acronym VOLE stands for Vector Oblivious Linear Evaluation, an interactive cryptographic technique that supplies two parties with correlated random values. These mathematical correlations function as internal authentication tags, empowering a verifier to check whether a computation performed on a hidden witness was executed correctly. In the VOLE-in-the-Head paradigm, the prover locally simulates the computational steps that would normally be executed interactively between two separate parties—hence the conceptual framing "in the head"—before committing to those values and applying a Fiat-Shamir transformation to convert the interactive process into a non-interactive zero-knowledge proof.

While the VOLE-in-the-Head proof system is not asymptotically succinct in the manner of modern zk-SNARKs, it is concretely efficient when applied to specific types of computational statements, particularly those that can be expressed as a series of algebraic equations defined over small binary fields. This algebraic suitability extends naturally to multivariate quadratic problems, where signers prove solutions to systems of random quadratic equations across multiple variables. Mathematical problems of this nature are widely believed to remain exceptionally difficult to solve, even for advanced quantum computers. Because these multivariate structures are constructed from quadratic equations, proving knowledge of a valid solution via VOLE-in-the-Head techniques remains computationally economical.

Multivariate Signatures and Emerging Blind Signature Schemes

An alternative approach to constructing post-quantum signatures involves embedding specialized algebraic structures directly into multivariate quadratic maps. In the Unbalanced Oil and Vinegar signature scheme, which is also currently being evaluated by NIST, the public key is represented by a multivariate quadratic map, while a valid signature consists of a vector satisfying a specific target vector derived from hashing and salting the message intended for signing.

The secret key within a UOV scheme functions as a mathematical trapdoor that grants the authorized signer a method for deriving valid solutions to the underlying quadratic equations. This trapdoor reveals a hidden partitioning of variables within each equation into distinct categories known as oil variables and vinegar variables. Because oil variables are never multiplied directly against one another, fixing the values of the vinegar variables transforms the complex quadratic system into a straightforward linear system that can be solved efficiently. Within the public key itself, these variables are thoroughly mixed to conceal the underlying structural trapdoor from external observers.

UOV boasts some of the smallest digital signature sizes among NIST post-quantum candidates, measuring just 135 bytes. However, this efficiency comes at the direct expense of massive public keys, which measure approximately 46.6 kilobytes. Alternative multivariate candidates like MAYO achieve significantly smaller public keys at the cost of moderately larger signatures. Because the arithmetic complexity required to evaluate a UOV or MAYO signature is directly comparable to evaluating standard multivariate quadratic instances, they pair exceptionally well with VOLE-in-the-Head methodologies.

Building upon this cryptographic foundation, recent academic literature introduced PoMFRIT, a specialized post-quantum blind signature scheme constructed by combining VOLE-in-the-Head techniques with multivariate signature frameworks. PoMFRIT achieved notable efficiency milestones for candidate post-quantum constructions, requiring modest signature sizes during redemption and minimal overhead during issuance. By substituting MAYO with UOV within the construction, engineers can drive signature sizes down further, trading larger public keys for performance gains in application environments where public keys can be effectively cached and reused.

This architectural flexibility means that the underlying building blocks of PoMFRIT can be seamlessly reconfigured to support every credential system envisioned for MOLE, including advanced endorsements, anonymous credit tokens, and rate-limited credentials. Across these diverse constructions, the primary computational cost is dictated by the size of underlying cryptographic hash evaluations, specifically the Keccak permutation utilized within SHAKE and TurboSHAKE. Because Keccak’s bitwise operations translate natively into the constraints required by VOLE-in-the-Head proofs, cryptographic engineers can systematically optimize performance by minimizing hash requirements and refining state parameters.

Evaluating the Path Forward for Web Privacy

As security researchers continue to refine these post-quantum primitives, current implementations of schemes like VOLE-IHAT for endorsements and VOLE-ACT for credentials demonstrate that practical post-quantum privacy engineering is rapidly moving from theoretical design into concrete protocol engineering. While the resulting data sizes are undeniably larger than their classical counterparts—reflecting the broader reality that post-quantum cryptography carries a heavy bandwidth footprint—runtime performance metrics remain highly encouraging, with core operations executing in just a few milliseconds.

These performance indicators suggest that the cryptographic community has reached a vital transition milestone. The primary challenge is no longer inventing functional post-quantum anonymous credential schemes, but rather engineering protocols that can absorb and amortize these increased computational and bandwidth costs efficiently across real-world web traffic. Just as classical blind signatures were historically cheap enough to redeem on every individual HTTP request, future web infrastructure will require intelligent system designs that amortize single redemptions across multiple user interactions.

Significant work remains before these protocols see widespread deployment across global edge networks. Security engineers must rigorously verify protocol soundness through comprehensive cryptographic proofs ensuring security as long as underlying multivariate assumptions hold true. Additionally, the broader cryptographic community continues to monitor ongoing cryptanalysis of multivariate schemes to ensure parameter stability. Once these security parameters fully mature, multivariate cryptography is poised to anchor a vast ecosystem of privacy-preserving web applications, securing user anonymity well into the quantum age.