Date: July 5, 2025
This document provides an analysis of the performance benchmarks for the Quasor AEAD scheme. The tests were conducted after the v6.1 performance optimizations, which significantly improved throughput by optimizing the internal duplexing loop.
The results show a major performance increase across all operations, confirming the success of the optimization efforts while retaining all of Quasor’s high-security features.
This test compares the encryption speed of the optimized Quasor against two widely used, hardware-accelerated AEADs: AES-256-GCM and ChaCha20-Poly1305.
| Cipher | Average Time (128KB) | Relative Speed |
|---|---|---|
| Quasor (v6.1) | ~0.89 ms | 1x (baseline) |
| AES-256-GCM | ~0.12 ms | ~7.7x faster |
| ChaCha20-Poly1305 | ~0.12 ms | ~7.7x faster |
The v6.1 optimizations yielded a ~7.3x performance increase for Quasor. While still slower than hardware-accelerated ciphers (an expected trade-off for its advanced security features), the performance gap has narrowed considerably from the previous version.
This benchmark measures the time required to derive a master key from a password and salt.
| Operation | Average Time |
|---|---|
Quasor::new() |
~24.53 ms |
The key derivation time remains in the tens of milliseconds, which is a critical security feature. Argon2id is an intentionally slow, memory-hard function designed to make offline brute-force attacks computationally infeasible.
This benchmark measures the isolated performance of the BLAKE3-based SIV nonce derivation.
| Size | Average Throughput |
|---|---|
| 1 KB | ~654 MiB/s |
| 128 KB | ~2.61 GiB/s |
| 2 MB | ~174 MiB/s |
With the benchmark corrected to isolate the derive_nonce function, the results confirm that the SIV mechanism is exceptionally fast and is not a performance bottleneck. The multi-gigabyte-per-second throughput for common message sizes demonstrates the efficiency of using BLAKE3 for this purpose.
This new benchmark group measures performance for common network packet sizes.
| Operation | Size | Average Throughput |
|---|---|---|
| Encryption | 64 bytes | ~37 MiB/s |
| 256 bytes | ~88 MiB/s | |
| 1500 bytes | ~111 MiB/s | |
| Decryption | 1500 bytes | ~108 MiB/s |
Quasor demonstrates excellent performance for small, latency-sensitive messages. The high throughput for typical packet sizes (e.g., 1500 bytes for Ethernet) makes it well-suited for securing network protocols.
These tests measure the sustained throughput of the optimized Quasor for large messages.
| Operation | Size | Average Throughput (v6.1) | Improvement vs. v6.0 |
|---|---|---|---|
| Encryption | 1 KB | ~106 MiB/s | ~6.0x |
| 128 KB | ~142 MiB/s | ~7.2x | |
| 1 MB | ~64 MiB/s | ~3.7x | |
| 4 MB | ~101 MiB/s | ~4.5x | |
| Decryption | 128 KB | ~138 MiB/s | ~4.6x |
| 4 MB | ~98 MiB/s | ~3.1x |
The v6.1 optimizations were highly effective. Quasor now offers substantially higher performance, making it more practical for a wider range of applications, while retaining its advanced, high-security design. The benchmarks validate that the trade-offs between speed and security are well-balanced, and the underlying cryptographic primitives are used efficiently.