Polar Codes for Module-LWE Public Key Encryption: The Case of Kyber

• Published in: Cryptography Vol. 7; no. 1; p. 2
• Main Authors: Papadopoulos, Iason, Wang, Jiabo
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.01.2023
• Subjects: Algorithms; Codes; Computers; Cryptography; Data encryption; Data security; Data transmission; decryption failure rate; Encryption; Error analysis; Error correcting codes; Error correction; Error correction & detection; Failure rates; Information transfer; Kyber; Methods; module LWE; Modules; Parameters; Performance evaluation; Polar Codes; post-quantum security; Public key encryption; Quantum computers; Quantum computing; Security; United Kingdom
• ISSN: 2410-387X; 2410-387X
• DOI: 10.3390/cryptography7010002

In modern society, the Internet is one of the most used means of communication. Thus, secure information transfer is inevitably of major importance. Computers nowadays use encryption methods based on arithmetic operations to turn messages into ciphertexts that are practically impossible for an attacker to reverse-engineer using a classical computer. Lately, it has been proven that this is possible in a post-quantum setting where quantum computers of considerable size are available to attackers. With the advance of technology of quantum computers, it is now more necessary than ever before to construct encryption schemes that cannot be broken either using a classical or a quantum computer. The National Institute of Technology and Standards (NIST) has orchestrated a competition, and numerous encryption schemes have been proposed. The NIST has identified one algorithm to be standardized for the post-quantum era. This algorithm is called CRYSTALS-Kyber and is based on module learning with errors (MLWE). This paper investigates how to apply error correcting codes in order to create some excess decryption failure rate (DFR) and to take advantage of that in order to re-tune Kyber’s parameters in the pursuit of higher security. By applying Polar Codes, Kyber’s security was managed to be increased by 54.4% under a new set of parameters, while keeping the decryption failure rate well below the upper acceptable bound set by the NIST.