Paper 2025/1421

Efficient randomized strong $2$-source non-malleable extractor for any linear min-entropy

Divesh Aggarwal, National University of Singapore
Pranjal Dutta, Nanyang Technological University
Saswata Mukherjee, National University of Singapore
Satyajeet Nagargoje, Georgetown University
Maciej Obremski, Georgetown University
Abstract

Randomness is a fundamental requirement in cryptographic systems, enabling secure encryption, commitments, and zero-knowledge proofs. However, real-world randomness sources often suffer from weaknesses that adversaries can exploit, leading to significant security vulnerabilities. While deterministic randomness extraction from a single min-entropy source is impossible, two-source extractors provide a robust solution by generating nearly uniform randomness from two independent weak sources. Moreover, cryptographic systems must also be resilient to leakage and tampering attacks, necessitating the development of non-malleable two-source extractors. In this work, we construct a two-source non-malleable extractor in the Common Reference String (CRS) model, where a random low-degree polynomial is sampled once and made accessible to independent random sources, the distinguisher, and the tamperer. Our extractor requires only linear min-entropy in both sources and doesn't rely on strong computational assumptions, in contrast to prior constructions requiring computational assumptions such as sub-exponential hardness of the Decisional Diffie-Hellman (DDH) problem. Notably, our construction builds upon and relies on the recent breakthrough proof of the polynomial Freiman-Ruzsa conjecture. A connection of the Freiman-Ruzsa conjecture with two-source extractors was considered in prior work [ZBS11],[AGMR24], but their construction did not achieve non-malleability. Our results advance the state of non-malleable cryptographic primitives, with applications in secure storage, leakage-resilient cryptography, and privacy amplification. By eliminating the need for strong computational hardness assumptions, our techniques provide a more foundational and widely applicable method for randomness extraction. We also show, that the requirements on CRS for our application are so mild that the CRS can be sampled with $2$ party computation even when one of the parties is malicious (setting in which establishing unbiased coins is impossible).

Metadata
Available format(s)
PDF
Category
Foundations
Publication info
A minor revision of an IACR publication in CRYPTO 2025
Keywords
Randomness ExtractionTamper-resilient cryptographyLeakage-resilient cryptography
Contact author(s)
divesh aggarwal @ gmail com
pranjal dutta @ ntu edu sg
saswatamukherjee607 @ gmail com
satyajeetn2012 @ gmail com
obremski math @ gmail com
History
2025-08-04: approved
2025-08-04: received
See all versions
Short URL
https://ia.cr/2025/1421
License
Creative Commons Attribution
CC BY

BibTeX

@misc{cryptoeprint:2025/1421,
      author = {Divesh Aggarwal and Pranjal Dutta and Saswata Mukherjee and Satyajeet Nagargoje and Maciej Obremski},
      title = {Efficient randomized strong $2$-source non-malleable extractor for any linear min-entropy},
      howpublished = {Cryptology {ePrint} Archive, Paper 2025/1421},
      year = {2025},
      url = {https://eprint.iacr.org/2025/1421}
}
Note: In order to protect the privacy of readers, eprint.iacr.org does not use cookies or embedded third party content.