Security & Cryptography Benchmarks

Detailed performance metrics for the Nalix security primitives, including encryption ciphers, handshake computation, and hashing algorithms.

Envelope Cipher Suites#

Comparison of stream ciphers (Salsa20 and ChaCha20) and AEAD suites (Salsa20-Poly1305 and ChaCha20-Poly1305) over 64 B and 1024 B payloads. All operations are run on zero-allocation spans.

Envelope Cipher Performance (Payload Size = 64)#

Method	PayloadSize	Mean	Error	StdDev	P95	Allocated
Encrypt_Salsa20	64	331.0 ns	10.13 ns	11.66 ns	342.7 ns	0 B
Decrypt_Salsa20	64	260.2 ns	5.27 ns	6.07 ns	265.0 ns	0 B
Encrypt_Chacha20	64	343.4 ns	4.49 ns	5.17 ns	347.8 ns	0 B
Decrypt_Chacha20	64	260.1 ns	6.18 ns	7.12 ns	263.5 ns	0 B
Encrypt_Salsa20Poly1305	64	734.6 ns	27.18 ns	31.30 ns	752.7 ns	0 B
Decrypt_Salsa20Poly1305	64	712.0 ns	95.41 ns	109.87 ns	773.4 ns	0 B
Encrypt_Chacha20Poly1305	64	511.2 ns	5.81 ns	6.69 ns	520.5 ns	0 B
Decrypt_Chacha20Poly1305	64	484.0 ns	3.58 ns	4.13 ns	492.0 ns	0 B

Envelope Cipher Performance (Payload Size = 1024)#

Method	PayloadSize	Mean	Error	StdDev	P95	Allocated
Encrypt_Salsa20	1024	1,895.6 ns	25.27 ns	28.09 ns	1,940.0 ns	0 B
Decrypt_Salsa20	1024	1,887.3 ns	21.79 ns	25.09 ns	1,928.4 ns	0 B
Encrypt_Chacha20	1024	1,127.3 ns	265.28 ns	305.50 ns	1,419.2 ns	0 B
Decrypt_Chacha20	1024	1,301.6 ns	59.30 ns	68.29 ns	1,334.3 ns	0 B
Encrypt_Salsa20Poly1305	1024	4,526.5 ns	64.45 ns	74.22 ns	4,583.9 ns	0 B
Decrypt_Salsa20Poly1305	1024	4,606.4 ns	12.49 ns	14.39 ns	4,623.0 ns	0 B
Encrypt_Chacha20Poly1305	1024	2,695.9 ns	59.62 ns	68.66 ns	2,743.3 ns	0 B
Decrypt_Chacha20Poly1305	1024	2,736.3 ns	60.69 ns	69.89 ns	2,788.0 ns	0 B

Behind the design#

• Software-Efficient Stream Ciphers: Salsa20 and ChaCha20 perform encryption at near-hardware speeds. The state is maintained in stackalloc memory, ensuring zero allocations.
• One-Pass AEAD: The combined ciphers with Poly1305 perform authentication and decryption in a single pass over the memory block, minimizing cache misses.

---

Handshake Protocol#

Performance metrics for the cryptographic handshake phase, establishing session identity.

Method	Mean	Error	StdDev	P95	Allocated
ComputeMasterSecret	2.513 μs	0.1719 μs	0.1980 μs	2.614 μs	0 B
ComputeServerProof	2.194 μs	0.2906 μs	0.3346 μs	2.609 μs	0 B
ComputeClientProof	2.368 μs	0.3165 μs	0.3645 μs	2.654 μs	0 B
DeriveSessionKey	2.619 μs	0.0755 μs	0.0869 μs	2.675 μs	0 B

Behind the design#

• Zero-Allocation Keys: High-cost key computations take ~2–2.6 μs each and run completely allocation-free (0 B), isolating cryptographic handshakes from garbage collection impacts.

---

Hashing & Cryptography#

Comparison of cryptographic hashing, verification ciphers, and key derivation algorithms.

Hashing Performance (Payload Size = 64)#

Method	PayloadSize	Mean	Error	StdDev	P95	Allocated
Keccak256_Hash	64	605.3 ns	11.16 ns	12.85 ns	615.0 ns	0 B
HmacKeccak256_Compute	64	2,518.4 ns	45.27 ns	52.13 ns	2,553.1 ns	0 B
Poly1305_Compute	64	111.5 ns	0.95 ns	1.09 ns	112.3 ns	0 B
Pbkdf2_Hash	64	2,561,666.8 ns	43,716.23 ns	50,343.68 ns	2,609,432.4 ns	312 B
Pbkdf2_Verify	64	2,548,772.0 ns	39,314.36 ns	45,274.48 ns	2,590,398.3 ns	256 B

Hashing Performance (Payload Size = 1024)#

Method	PayloadSize	Mean	Error	StdDev	P95	Allocated
Keccak256_Hash	1024	4,862.0 ns	64.07 ns	71.22 ns	4,967.1 ns	0 B
HmacKeccak256_Compute	1024	6,591.5 ns	140.27 ns	161.54 ns	6,758.3 ns	0 B
Poly1305_Compute	1024	888.9 ns	57.31 ns	66.00 ns	940.5 ns	0 B
Pbkdf2_Hash	1024	2,545,073.2 ns	49,667.27 ns	57,196.90 ns	2,608,114.0 ns	312 B
Pbkdf2_Verify	1024	2,552,527.7 ns	49,817.62 ns	57,370.04 ns	2,601,772.5 ns	256 B

Behind the design#

• Keccak Optimization: Our custom Keccak256 implementation hashes small structures in ~605 ns and 1KB structures in 4.86 μs without allocating memory.
• PBKDF2 Overhead & Isolation: PBKDF2 is a secure key derivation function designed to be intentionally computationally expensive (taking ~2.5 ms per operation). Due to its CPU-bound nature, PBKDF2 checks are isolated from the networking hot path to prevent starvation of the listener thread pool.