A Constant-Rate Compiler for Continual Memory Leakage

What it is

ePrint 2026/2715 (received 1 November 2026, approved 4 November 2026) is a preprint by Shweta Agrawal, Anuja Modi and Rajeev Raghunath. The page notes "To be submitted", so it has not been peer reviewed. It gives a compiler for the continual memory leakage (CML) model with constant leakage rate. All earlier general compilers in this model with two independently leaking components tolerated only an inverse-polynomial rate. The authors say at least two components are necessary.

How it works

The compiler works in a hybrid model with a simple inputless leak-free sampler, as in Dziembowski and Faust (TCC 2012). It packs the whole configuration of a universal machine into one element of a large extension field. At any moment only a constant number of inner-product encodings sit in memory, each for a constant number of rounds. The authors say the idea is adapted from packing techniques in homomorphic encryption.

The compiler uses no computational assumption, so the leakage protection itself is information-theoretic. Computational assumptions enter only through the base scheme being compiled.

Preconditions / Affected

• The compiled computation must be deterministic, on a fixed secret, with public inputs and outputs.
• The leak-free sampler is required.
• A transfer theorem covers any security game in which the adversary has an oracle for the computation. Replacing each oracle call with a compiled evaluation that leaks continuously keeps the game secure.

Applications

Applied to the decryption of any IND-CCA2 PKE, the compiler gives IND-CCLA2 security. This means CCA2 security even when every decryption leaks, including after the challenge ciphertext. The public key, ciphertexts and encryption algorithm stay unchanged, so only the decryption side is replaced. Instantiating the base scheme gives such PKE from LWE and from LPN. The authors say these improve on the previous best post-quantum constructions in this model, which supported only inverse-polynomial rates. A natural generalization gives constant-rate signatures, PRFs and MACs.

Caveats

The page gives no concrete leakage-rate constant, no overhead figures and no implementation. Whether it is practical is not addressed.