On the Efficiency of Software Implementations of Lightweight Block Ciphers from the Perspective of Programming Languages

This repository contains the implementation codes from the paper "On the Efficiency of Software Implementations of Lightweight Block Ciphers from the Perspective of Programming Languages".

Key Contributions

• Analysis of Implementation Choices: We examine how different software implementations, programming languages, and platforms (including cloud environments) affect the performance of the LED block cipher.
• Implementation Types: We define six different lookup-table-based (LUT) software implementations for the LED cipher, ranging from memory-optimized to throughput-optimized variants.
• Platform Comparison: We benchmark 72 unique implementations (combinations of four programming languages and three key sizes) across 32-bit and 64-bit platforms running two major operating systems (Windows and Linux) on Amazon Web Services (AWS) cloud.
• Results: Our benchmarks reveal significant performance differences across platforms and languages, highlighting the impact on throughput and software efficiency.

Paper Link

The full paper is available for download are ScienceDirect and ePrint

Implementations

We implemented the LED block cipher for three different key sizes (64, 80, and 128 bits) using six different LUT-based methods. These implementations were coded in the following languages:

• Java
• C#
• C++
• Python

The implementation types include:

1. 4-bit Serial
2. 4-bit LUT
3. 8-bit LUT
4. 16-bit LUT
5. 32-bit LUT
6. 64-bit LUT

Each method was designed with trade-offs between memory consumption and throughput performance, offering various optimizations depending on the use case.

Benchmarking

We performed thorough benchmarking using EC2 instances in the AWS cloud, running the implementations on both 32-bit and 64-bit versions of Windows and Linux. The following performance metrics were considered:

• Throughput: The rate at which data is encrypted or decrypted.
• Software Efficiency (SE): The throughput per unit of code size, indicating the efficiency of the implementation.