libmc/docs/Embedded-cryptography.md

Starting with version 0.2.0 (26.2-0.2.0), libmc includes a built-in cryptography implementation. It uses the JDK's built-in cryptographic APIs on the JVM platform and a pure-Kotlin implementation on kotlin-native platforms. However, due to a lack of low-level optimizations, performance results in encoding and decoding tests are significantly slower than JDK's built-in implementations

A simplified table below shows the benchmark results:

Algorithm	Data Size (Bytes)	libmc's Built-in (ops/s)	JDK's Built-in (ops/s)	Gap
AES-128-CFB8 Encrypt	32	112,554.27	1,527,235.11	~13.6 times slower than JDK's built-in
	1,024	2,976.95	49,842.94	~16.7 times slower than JDK's built-in
	65,536	45.04	773.46	~17.2 times slower than JDK's built-in
AES-128-CFB8 Decrypt	32	115,031.69	1,515,218.15	~13.2 times slower than JDK's built-in
	1,024	2,925.74	49,868.55	~17.0 times slower than JDK's built-in
	65,536	48.47	777.11	~16.0 times slower than JDK's built-in
RSA-1024 Encrypt	16	163.87	91,600.51	~559 times slower than JDK's built-in
	64	158.27	93,042.81	~587 times slower than JDK's built-in
	117	160.95	94,826.94	~589 times slower than JDK's built-in
SHA-1 Hashing	64	2,134,974.72	11,609,497.92	~5.4 times slower than JDK's built-in
	1,024	259,804.80	1,856,575.14	~7.1 times slower than JDK's built-in
	65,536	4,485.95	32,132.67	~7.2 times slower than JDK's built-in

Benchmark Environment:

• CPU: AMD Ryzen 5 5500U (6 Cores / 12 Threads @ 2.10GHz)
• RAM: 16GB DDR4 2667MHz
• OS: Windows 11 64-bit
• Runtime: Microsoft Build of OpenJDK 17.0.8, Kotlin 2.4.10 (MingwX64 & JVM)

Fortunately, except for AES-128-CFB8 (which requires continuous stream encryption/decryption during networking), the other operations (RSA & SHA-1) are only executed once during the initial server authentication phase

If you have optimized native algorithm implementations (via cinterop or other approaches), PRs are welcome