Toshihiko Okamura

A Hybrid ARQ Scheme Based on Shortened Low-Density Parity-Check Codes

This paper presents a hybrid ARQ (HARQ) scheme based on shortened low-density parity-check (LDPC) codes. Although conventional incremental redundancy HARQ (IR-HARQ) schemes based on punctured LDPC codes can achieve high throughput, puncturing slows the convergence of iterative decoding, and moreover, LDPC codes originally designed to be high-rate generally have lower decoding complexity for each iteration than low-rate codes. This makes computational complexity at high rates a problem for conventional IR-HARQ schemes. The proposed HARQ scheme uses a high-rate mother code: for a retransmission request, information bits are divided into smaller sub-blocks, and additional parity bits are generated for one of the sub-blocks to be used as a shortened form of the mother code. In order to achieve high throughput with the proposed HARQ scheme, it is necessary for the shortened codes of all the sub-blocks to perform well. A multi-edge type code-design is employed to construct irregular LDPC codes that meet this requirement. Simulation results show that the proposed HARQ scheme can achieve throughput comparable to that of a conventional IR-HAQR scheme over a wide range of signal-to-noise ratios and with lower computational complexity.

Designing LDPC codes using cyclic shifts

This paper presents a method to construct LDPC codes by arranging cyclic shift matrices. These codes perform well at various rates for moderate code length. By substituting those cyclic shifts for their Kronecker products with small random permutations, the performance of long codes can be improved, while the structure facilitates hardware implementation.

A smaller and faster variant of RSM

Masking is one of the major countermeasures against side-channel attacks to cryptographic modules. Nassar et al. recently proposed a highly efficient masking method, called Rotating S-boxes Masking (RSM), which can be applied to a block cipher based on Substitution-Permutation Network. It arranges multiple masked S-boxes in parallel, which are rotated in each round. This rotation requires remasking process for each round to adjust current masks to those of the S-boxes. In this paper, we propose a method for reducing the complexity of RSM further by omitting the remasking process when the linear diffusion layer of the encryption algorithm has a certain algebraic property. Our method can be applied to AES with a reduced complexity from RSM, while keeping the equivalent security level.