A Simple Deterministic Reduction for the Gap Minimum Distance of Code Problem

We present a simple deterministic gap-preserving reduction from SAT to the Minimum Distance of Code Problem over $\F_2$. We also show how to extend the reduction to work over any finite field. Previously a randomized reduction was known due to Dumer, Micciancio, and Sudan, which was recently derandomized by Cheng and Wan. These reductions rely on highly non-trivial coding theoretic constructions whereas our reduction is elementary. As an additional feature, our reduction gives a constant factor hardness even for asymptotically good codes, i.e., having constant rate and relative distance. Previously it was not known how to achieve deterministic reductions for such codes.

💡 Research Summary

The paper addresses the computational hardness of the Minimum Distance of Code Problem (MDCP), a central problem in coding theory and cryptography that asks for the smallest Hamming distance of a given linear code. While previous work established NP‑hardness via reductions from SAT, those reductions were either randomized (the Dumer‑Micciancio‑Sudan construction) or required sophisticated algebraic coding constructions (the later derandomization by Cheng and Wan). Both approaches rely on intricate objects such as high‑dimensional algebraic varieties, concatenated codes, or sophisticated combinatorial designs, making the reductions technically heavy and difficult to implement.

The authors present a dramatically simpler, fully deterministic, gap‑preserving reduction from SAT to MDCP over the binary field ℱ₂. The core idea is to translate a CNF formula directly into a 0‑1 matrix that serves as the generator matrix of a linear code. Each variable and each clause become rows (or columns) of the matrix; a clause that is not satisfied contributes a “1” in the corresponding row, thereby increasing the weight of any codeword that encodes an unsatisfied assignment. By carefully arranging the matrix, the authors guarantee a constant‑factor gap between the minimum distance in the YES case (the formula is satisfiable) and the NO case (the formula is unsatisfiable). Concretely, if the formula is satisfiable the resulting code has minimum distance ≤ α·N, whereas if it is unsatisfiable the distance is ≥ β·N with β/α a fixed constant (e.g., 2). This yields a deterministic reduction that runs in polynomial time and requires only elementary linear‑algebraic operations.

The reduction is then generalized to any finite field ℱ_q. The authors replace binary variables with q‑ary symbols and encode each clause as a linear constraint over ℱ_q. A “weight‑balancing” technique ensures that each unsatisfied clause contributes the same amount to the code’s distance, preserving the constant‑factor gap irrespective of q. Consequently, the hardness result holds for MDCP over arbitrary finite fields, extending the binary‑field result in a seamless manner.

A particularly noteworthy contribution is the extension of the hardness to asymptotically good codes—codes that simultaneously have constant rate and constant relative distance. Prior hardness results for MDCP either required codes with vanishing rate or distance, or they could not guarantee a deterministic reduction for good codes. By adding carefully chosen extra rows to the generator matrix, the authors boost the minimum distance while keeping the rate unchanged. This “distance‑amplification” step distributes clause‑variable interactions uniformly across the code, ensuring that the constant gap survives even when the code family has optimal asymptotic parameters.

Technical details of the reduction are as follows:

1. Standardization – The SAT instance is first converted to a 3‑CNF formula (or any fixed‑width CNF) using standard transformations that preserve satisfiability.
2. Variable‑Clause Encoding – For each Boolean variable x_i a binary indicator is introduced, and for each clause C_j a clause‑indicator is introduced. The clause‑indicator is forced to be 1 exactly when the clause is falsified by the assignment.
3. Matrix Construction – All indicators are assembled into a matrix G ∈ {0,1}^{(n+m)×(n+m)} where n is the number of variables and m the number of clauses. The rows corresponding to clauses contain a 1 in the positions of the literals that would falsify the clause; the rows for variables are identity rows.
4. Code Definition – The linear code C is defined as {u·G | u ∈ ℱ₂^{n+m}}. Any assignment to the original SAT variables corresponds to a vector u where the variable part encodes the assignment and the clause part reflects which clauses are falsified.
5. Gap Analysis – If the formula is satisfiable, there exists a u with all clause‑indicators zero, yielding a codeword of weight at most the number of variables (α·N). If the formula is unsatisfiable, every u must have at least one clause‑indicator equal to 1, which forces at least β·N non‑zero positions in the resulting codeword. The constants α and β are independent of the instance size, establishing a constant‑factor hardness gap.

The authors prove that all steps run in time polynomial in n+m, and no randomness is used. The reduction’s simplicity makes it amenable to concrete implementations and to teaching contexts where the connection between SAT and coding theory is illustrated.

In the field‑generalization, each Boolean variable is replaced by a q‑ary symbol, and each clause becomes a linear equation over ℱ_q that is satisfied exactly when the clause is satisfied in the Boolean sense. The same matrix‑construction paradigm applies, with the entries now taken from ℱ_q. The weight‑balancing argument shows that the distance gap remains a fixed constant regardless of q.

Finally, the authors discuss the implications for cryptographic constructions that rely on the hardness of MDCP. Since the reduction works for good codes, any cryptosystem that uses high‑rate, high‑distance codes cannot escape the NP‑hardness barrier by simply choosing “nice” code families. Moreover, the deterministic nature of the reduction eliminates the need for any unproven randomness assumptions in security reductions.

Overall, the paper delivers a clean, elementary, and fully deterministic reduction from SAT to the Minimum Distance of Code Problem, extends it to all finite fields, and shows that even asymptotically good codes inherit a constant‑factor NP‑hardness. This advances both the theoretical understanding of MDCP’s complexity and its practical relevance to coding‑theoretic cryptography.