New method shows some digital security math is easier to solve
Researchers found that certain mathematical problems used in cryptography are less secure than previously estimated.
The Regular Syndrome Decoding (RSD) problem is easier to solve than experts thought. Many cryptographic constructions rely on the hardness of RSD. These include zero knowledge protocols and pseudorandom correlation generators.
Researchers used a refined hybrid algebraic attack to solve these problems. This method coordinates zero coordinate guesses and syndrome based variable eliminations. A guessing stage involves making choices to reduce the problem. Linear elimination is a process used to solve equations.
The new method uses a block balancing principle. It coordinates the guessing and elimination stages across all regular blocks. This balances the guessing penalty while keeping enough equations for linearization.
This work reassesses applications like Ferret, Wolverine, and committed VOLE. At least one parameter set per application falls below the 128 bit target. Some losses approach 67 bits.
On PCG benchmarks, attack costs improved by up to 29.77 and 25.91 bits. For revised 128 bit PCG parameter sets, estimated attack costs are all below field operations. Existing parameter choices may overestimate the hardness of RSD instances.
Key facts
- At least one parameter set per application falls below its 128 bit target.
- Attack costs on PCG benchmarks improved by up to 29.77 and 25.91 bits.
- Revised 128 bit PCG parameter set attack costs are below field operations.