isap-lwc / isap-code-package

Code package for ISAP, a family of lightweight authenticated encryption algorithms designed with a focus on robustness against implementation attacks that is currently competing in the final round of NIST's standardization effort for lightweight authenticated encryption schemes.

ISAP Code Package

ISAP is a family of lightweight authenticated encryption algorithms designed with a focus on robustness against implementation attacks and reached the final round of the NIST Standardization Process for Lightweight Cryptography (2019-2023). ISAP is of particular interest for applications like firmware updates where robustness against power analysis and fault attacks is crucial while code size and a small footprint in hardware matters. ISAP's original version was published at FSE 2017.

This repository contains implementations of the following algorithms:

• crypto_aead/isapa128av20: ISAP-A-128a (primary recommendation)
• crypto_aead/isapa128v20: ISAP-A-128
• crypto_aead/isapk128av20: ISAP-K-128a
• crypto_aead/isapk128v20: ISAP-K-128
• crypto_aead_hash/isapa128av20: ISAP-A-128a + Ascon-Hash

and the following implementations:

• avx_512: AVX-512 implementation in C.
• opt_64: Optimized 64-bit implementation in C.
• opt_64_compact: More compact variant of opt_64.
• opt_64_stp: Variant of opt_64_compact that includes a leakage-resilient tag comparison.
• opt_32: Optimized 32-bit implementation in C.
• opt_32_compact: More compact variant of opt_32.
• opt_32_tiny: Much more compact variant of opt_32.
• opt_32_stp: Variant of opt_32_compact that includes a leakage-resilient tag comparison.
• bi_32: Bit-interleaved 32-bit implementation in C.
  • Should perform better than opt_32 on 32-bit processors that feature a rotation instruction such as the ARM Cortex-M3/4 or the Raspberry Pi.
• bi_32_compact: More compact variant of bi_32.
• bi_32_stp: Variant of bi_32_compact that includes a leakage-resilient tag comparison.
• bi_32_armv6: Bit-interleaved 32-bit implementation using ARMv6 assembly.
• bi_32_armv6m: Bit-interleaved 32-bit implementation using ARMv6m assembly.
• bi_32_armv6m_compact: More compact variant of bi_32_armv6m.
• bi_32_armv6m_stp: Variant of bi_32_armv6m_compact that includes a leakage-resilient tag comparison.
• bi_32_armv7m: Bit-interleaved 32-bit implementation using ARMv7m assembly.
• opt_8: Optimized 8-bit implementation in C.
• opt_8_tiny: More compact variant of opt_8.
• bi_8: Bit-interleaved 8-bit implementation in C.
  • Should perform better than opt_8 on 8-bit processors that feature a rotation instruction.
• ref: Easy-to-read implementation in C (not suitable for benchmarks).

A summary of the robustness features of the ISAP mode against implementation attacks can be found in docs/documentation.pdf.

Performance on different platforms (cycles/byte):

ISAP-A-128a (primary recommendation)

Message Length in Bytes:	64	1536	long
AMD EPYC 7742 (x64) ¹	-	-	18.9
AMD Ryzen 7 1700 (x64)	85.7	24.5	21.9
ARM Cortex-A53 (ARMv8) ¹	-	-	38.3
Raspberry Pi 1B (ARMv6)	551	171	159
STM32F303 (ARMv7m)	542	168	150
ATmega328P (AVR) ²	450	-	-

¹ Taken from eBACS.

² Taken from NIST LWC benchmarks.

For up-to-date benchmark results click here.

Run KATs:

Compile and test all implementations for ISAP-A-128a (primary recommendation):

1. cd tests
2. bash test_all.sh (ARM assembly variants are excluded by default)

Compile and test a specific implementation:

1. Use one of the compile commands provided below (adjust as needed)
2. Execute: ./genkat

x64: ISAP-A-128a

gcc -march=native -O3 -DNDEBUG -Icrypto_aead_hash/isapa128av20/opt_64 crypto_aead_hash/isapa128av20/opt_64/*.c -DCRYPTO_AEAD -Itests tests/genkat_aead.c -o genkat

x64: ISAP-A-128a + Ascon-Hash

gcc -march=native -O3 -DNDEBUG -Icrypto_aead_hash/isapa128av20/opt_64 crypto_aead_hash/isapa128av20/opt_64/*.c -DCRYPTO_AEAD -DCRYPTO_HASH -Itests tests/genkat_hash.c -o genkat

ARMv6: ISAP-A-128a

gcc -march=native -O3 -DNDEBUG -Icrypto_aead_hash/isapa128av20/bi_32_armv6 crypto_aead_hash/isapa128av20/bi_32_armv6/*.c -DCRYPTO_AEAD -Itests tests/genkat_aead.c -o genkat

ARMv6: ISAP-A-128a + Ascon-Hash

gcc -march=native -O3 -DNDEBUG -Icrypto_aead_hash/isapa128av20/bi_32_armv6 crypto_aead_hash/isapa128av20/bi_32_armv6/*.c -DCRYPTO_AEAD -DCRYPTO_HASH -Itests tests/genkat_hash.c -o genkat

Run Benchmarks:

Compile and benchmark all implementations for ISAP-A-128a (primary recommendation):

1. cd tests
2. bash bench_all.sh (ARM assembly variants are excluded by default)

Compile and benchmark a specific implementation:

1. Use one of the compile commands provided below (adjust as needed)
2. Execute: ./getcycles

Note: On ARM devices you likely need to first activate the performance monitor unit (PMU). For more information click here.

x64: ISAP-A-128a

gcc -march=native -O3 -DNDEBUG -Icrypto_aead_hash/isapa128av20/opt_64 crypto_aead_hash/isapa128av20/opt_64/*.c -DCRYPTO_AEAD -Itests tests/getcycles.c -o getcycles

x64: ISAP-A-128a + Ascon-Hash

gcc -march=native -O3 -DNDEBUG -Icrypto_aead_hash/isapa128av20/opt_64 crypto_aead_hash/isapa128av20/opt_64/*.c -DCRYPTO_AEAD -DCRYPTO_HASH -Itests tests/getcycles.c -o getcycles

ARMv6: ISAP-A-128a

gcc -march=native -O3 -DNDEBUG -Icrypto_aead_hash/isapa128av20/bi_32_armv6 crypto_aead_hash/isapa128av20/bi_32_armv6/*.c -DCRYPTO_AEAD -Itests tests/getcycles.c -o getcycles

ARMv6: ISAP-A-128a + Ascon-Hash

gcc -march=native -O3 -DNDEBUG -Icrypto_aead_hash/isapa128av20/bi_32_armv6 crypto_aead_hash/isapa128av20/bi_32_armv6/*.c -DCRYPTO_AEAD -DCRYPTO_HASH -Itests tests/getcycles.c -o getcycles

License Information:

See LICENSE.txt