Classifying Implementations of Cryptographic Primitives and Protocols that Use Post-Quantum Algorithms

Tushin Mallick, Cristina Nita-Rotaru, Ashish Kundu, Ramana Kompella · 2025

Read Classifying Implementations of Cryptographic Primitives and Protocols that Use Post-Quantum Algorithms by Tushin Mallick, Cristina Nita-Rotaru, Ashish Kundu, Ramana Kompella (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-17830.txt.

Introduction

Tushin Mallick and co-authors study in 2025 whether outside observers can tell which post-quantum security methods a system is using, just from its behavior. Such classification can help defense, but attackers can also use it to pick attacks that fit a target. Post-quantum methods use more memory and varied computing time, which may leave a fingerprint.

Tests span two code libraries, several secure protocols, and tools for a type of proof, across three operating systems. Observers could tell old and new methods apart with very high accuracy, and often name the exact method and library. Readers in security learn that new methods may leak their identity through timing and size.

Catalog record

CREATORTushin Mallick, Cristina Nita-Rotaru, Ashish Kundu, Ramana Kompella
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE70 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-17830

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Classifying Implementations of Cryptographic Primitives and Protocols that Use Post-Quantum Algorithms Tushin Mallick, Cristina Nita-Rotaru, Ashish Kundu, Ramana Kompella Abstract Classification techniques can be used to analyze system behaviors, network protocols, and cryptographic primitives based on identifiable traits. While useful for defensive purposes, such classification can also be leveraged by attackers to infer system configurations, detect vulnerabilities, and tailor specific attacks — including denial-of-service, key recovery, or downgrade attacks — against the categorized targets. In this paper, we study the feasibility of classifying post-quantum (PQ) algorithms by analyzing implementations of primitives such as key exchange and digital signatures, their usage within secure protocols, and integration into SNARK generation libraries. Unlike traditional cryptographic algorithms, PQ algorithms have larger memory needs and variable computation costs. Our research examines two post-quantum cryptography libraries, liboqs and CIRCL . We evaluate their implementations of TLS, SSH, QUIC, OpenVPN, and OpenID Connect (OIDC) across Windows, Ubuntu, and macOS. Additionally, we analyze two SNARK generation and verification libraries— pysnark and lattice_zksnark —on Ubuntu. Experimental results reveal that, depending on computational workload, (1) we can distinguish between classical and PQ key exchange and digital signature algorithms with accuracies of 98% and 100%, respectively; (2) we can determine the used algorithm within PQ key exchange or digital signature algorithms with 97% and 86% accuracy, respectively; (3) we can distinguish between implementations of the same algorithm in the liboqs and CIRCL libraries for key exchange and digital signature with 97% and 100% accuracy, respectively; and (4) within the same library – CIRCL we can distinguish the PQ and hybrid key exchange algorithm implementations with 97% accuracy. In the case of secure protocols, it is possible to (1) discern whether the key exchange is classical or PQ, and (2) identify the specific PQ key exchange algorithm employed. Finally, the SNARK generation and verification of pysnark and lattice_zksnark are distinguishable with an accuracy of 100%. We demonstrate the applicability of our classification methods to real systems by applying them to the Tranco dataset to identify domains that use PQ key exchange for TLS, and by integrating them in QUARTZ, an open source risk and threat analyzer created by CISCO. I Introduction Quantum computing leverages quantum mechanics to process data in fundamentally new ways, enabling exponentially faster solutions to complex problems than classical computers. A major area of impact is cryptography, as schemes like RSA and Diffie-Hellman rely on computational problems vulnerable to quantum attacks. Although practical quantum computers are not yet available, researchers are developing cryptographic methods resilient to future quantum threats and designed to prevent “harvest-now, decrypt-later” attacks. NIST conducted a multi-year initiative to evaluate and standardize cryptographic algorithms resistant to quantum attacks, leading to the establishment of post-quantum cryptography (PQC) [ 1 ] . The process concluded in 2023 with public drafts for encryption [ 2 ] and digital signatures [ 3 , 4 ] , followed by the release of the first PQC standards—FIPS 203 [ 5 ] , FIPS 204 [ 6 ] , and FIPS 205 [ 7 ] —in August 2024 [ 8 ] . In March 2025, NIST selected Hamming Quasi-Cyclic (HQC) as a fifth key encapsulation mechanism, providing a non-lattice alternative to ML-KEM (FIPS 203), with a draft standard expected in 2026 and finalization by 2027. Concurrently, draft NIST IR 8547, Transition to Post-Quantum Cryptography Standards, offers guidance on migration planning, risk assessment, crypto agility, and asset inventory, recommending deprecation of quantum-vulnerable algorithms (RSA, ECDSA, EdDSA, DH, ECDH) by 2030 and their full retirement by 2035. Implementations of post-quantum cryptographic (PQC) primitives are now publicly available in libraries and integrated into secure communication protocols and storage systems. Notable examples include liboqs [ 9 ] , CIRCL [ 10 ] , AWS-LC [ 11 ] , pqm4 [ 12 ] , PQClean [ 13 ] , and libpqcrypto [ 14 ] . liboqs , developed by the Open Quantum Safe project under the Linux Foundation’s Post-Quantum Cryptography Alliance, supports integration and evaluation of PQC algorithms across major operating systems. CIRCL , by Cloudflare, provides advanced cryptographic algorithms for the Go programming language. Amazon’s AWS-LC library incorporates PQC schemes to strengthen AWS security. The pqm4 library targets resource-constrained embedded systems on ARM Cortex-M4. PQClean offers clean, portable C implementations of NIST PQC candidates for reference and reuse, while libpqcrypto , from the PQCRYPTO project, collects diverse encryption and signature schemes for research and benchmarking. Additionally, it has become crucial to secure protocols like QUIC, SSH, DTLS [ 15 ] , TLS, and OAuth 2.0 against quantum threats. Major tech companies are already adopting hybrid and quantum-safe solutions: Google, Microsoft, Cloudflare, and Amazon integrate quantum algorithms into their protocols; IBM offers quantum-safe TLS; Meta uses hybrid key exchange; Akamai prepares PQC deployment across its CDN; and Cisco embeds quantum-safe methods in its networking products—signaling an industry-wide shift toward quantum-resilient infrastructure. Post-quantum algorithms have distinct memory footprints and CPU cycle patterns, due to their complex operations, larger keys, and ciphertexts. These characteristics make them identifiable through classification techniques that analyze behavioral patterns and protocol-specific responses. Such classification can expose systems to risks, as attackers may exploit identifiable traits to launch denial-of-service, key recovery, or downgrade attacks. Conversely, it can aid defenders in assessing risks and identifying vulnerable components. Previous research has focused on classifying secure protocols using classical cryptography, underscoring the growing importance of similar approaches for post-quantum systems. Our contribution. In this work we investigate whether post-quantum cryptographic primitives can be classified based on their library implementations, use in secure communication protocols, or integration in advanced constructions like SNARKs. The classification task is challenging because post-quantum algorithms exhibit diverse computational and memory characteristics—some requiring more resources, others less, and some closely matching classical algorithms. These inconsistencies make rule-based approaches unreliable, motivating the use of machine learning models to capture subtle behavioral differences between classical and post-quantum schemes. We start with publicly available post quantum libraries from Microsoft, Cloudflare, Amazon to show that, when compared to classical algorithms in terms of computational cost and memory overhead their post quantum counterparts are easily distinguishable by using machine learning models. We collected data from liboqs and CIRCL running on Windows, Ubuntu, and MacOS under a scenario when other intensive computation tasks share the resources with the post quantum libraries. We formulate distinguishing key exchange schemes as well as signature schemes, employing machine learning methods to carry out the classification. Ensemble learning models, particularly XGBoost, exhibit the highest performance across various classification tasks. Specifically, it achieved 98% accuracy in classifying classical vs. post quantum key exchange algorithms and 100% accuracy for classical vs. post-quantum digital signatures. For post-quantum key exchange algorithms, the classification accuracy reached 97%, while for post-quantum digital signatures, it was 86%. XGBoost also effectively distinguished between implementations of the same algorithm in the liboqs and CIRCL libraries, achieving 96% accuracy in key exchange and 100% in digital signature. For post-quantum and hybrid key exchange classification within CIRCL library, it achieved 97% accuracy. We next proceed to see if classification is possible when post quantum algorithms are implemented for key exchanges in widely used protocols such as TLS [ 16 ] , SSH [ 17 ] , QUIC [ 18 ] , OpenVPN [ 19 ] and Open ID

Read the complete text