Add SRAM PUF (Physically Unclonable Function) support to wolfCrypt. Derives device-unique cryptographic keys from the power-on state of SRAM memory using a BCH(127,64,t=10) fuzzy extractor with HKDF key derivation.
- **wolfCrypt PUF API** (`wolfcrypt/src/puf.c`, `wolfssl/wolfcrypt/puf.h`)
- `wc_PufInit`, `wc_PufReadSram`, `wc_PufEnroll`, `wc_PufReconstruct`
- `wc_PufDeriveKey` (HKDF-SHA256), `wc_PufGetIdentity` (SHA-256 device fingerprint)
- `wc_PufZeroize` (secure context cleanup)
- `wc_PufSetTestData` (synthetic SRAM for testing without hardware)
- **BCH(127,64,t=10) error-correcting codec** - corrects up to 10 bit flips per 127-bit codeword across 16 codewords
- **`WC_PUF_SHA3` build option** - select SHA3-256 instead of SHA-256 for identity hash and HKDF (default: SHA-256)
- **Precomputed GF(2^7) tables** - `const` arrays in `.rodata` (no runtime init, thread-safe, flash-resident on embedded)
- `./configure --enable-puf` (auto-enables HKDF dependency)
- CMake: `WOLFSSL_PUF=yes`
- `WOLFSSL_USER_SETTINGS`: define `WOLFSSL_PUF` and `WOLFSSL_PUF_SRAM`
- See wolfssl-examples/puf for example implementation on STM32 NUCLEO-H563ZI (Cortex-M33, STM32H563ZI)
- Supports test mode (synthetic SRAM)
- Builds to ~13KB `.elf`
- Tested on NUCLEO-H563ZI: enrollment, noisy reconstruction, key derivation all pass
- `.github/workflows/puf.yml`: host build + test workflow for PUF feature
- Doxygen API docs for all 8 public functions
- PUF group added to `doxygen_groups.h`
Refactors the output format of generated assembly files across all
platforms
(x86_64, ARM AArch64, ARMv8-32, Thumb2, PowerPC) for consistency and
correctness.
Changes
Data constant consolidation
- Pack multiple values per directive line (e.g., 4× .long or 8× .short
per
line) instead of one value per line, reducing file sizes significantly
- Normalize hex literal formatting: 64-bit values use full 8-byte
zero-padded
form (e.g., 0x0000000003ffffff instead of 0x3ffffff)
x86_64 assembly
- Use decimal immediate values for shift counts (e.g., $1 instead of
$0x01)
- .asm (MASM): use hex notation consistently for data constants;
update ALIGN
values to match data width (e.g., ALIGN 32 for 256-bit aligned data)
ARM .S files
- Move .type directive before .section for data objects (correct
ordering per
ELF convention)
ARM/Thumb2 inline C (_c.c) files
- Replace asm( with __asm__( for register variable constraints (better
portability)
- Add XALIGNED(8) attribute to constant lookup tables used in inline
asm
- Remove redundant #include <stdint.h> and
<wolfssl/wolfcrypt/libwolfssl_sources.h> headers
Files affected: 71 assembly and companion C files across
wolfcrypt/src/,
wolfcrypt/src/port/arm/, covering AES, ChaCha, Poly1305,
SHA-256/512/3,
Curve25519, ML-KEM, ML-DSA, and SP math routines.
1. sp_cond_swap_ct_ex (line ~5524) — XOR typo: b->sign ^= b->sign always
zeroed the sign. Fixed to b->sign ^= t->sign to correctly swap signs.
2. sp_mod_d (line ~7271) — Negative modulo correction was applied even
when the remainder was 0. Added (*r != 0) guard to avoid producing d
instead of 0.
3. sp_lshb (line ~8444) — Left-shift size check was off. Refactored to
correctly distinguish between pure-digit shifts and bit-within-digit
shifts when checking if the result fits, using separate overflow checks
for each case.
4. _sp_mulmod_tmp (line ~12160) — Zero inputs caused an allocation of
size 0, which is problematic. Added an early path: if either operand is
zero, set result to zero and skip the allocation/multiply entirely.
5. sp_mod_2d — copy path (line ~14762) — XMEMCPY copied digits *
SP_WORD_SIZEOF bytes but a may have fewer than digits used digits. Fixed
to copy min(a->used, digits) digits to avoid reading uninitialized
memory.
6. sp_mod_2d — negation loop (line ~14782) — Negation loop iterated
over r->used, which could exceed digits. Fixed to loop over min(r->used,
digits).
7. _sp_sqrmod (line ~17314) — Same zero-input issue as _sp_mulmod_tmp.
Added early zero path to skip the allocation/squaring when input is
zero.
8. sp_lcm (line ~19838) — Typo in sign check: b->sign >= MP_NEG
(comparing against a value that is 1, so >= 1 would also match MP_ZPOS)
changed to b->sign == MP_NEG.
Juliusz Sosinowicz
Sean Parkinson
JacobBarthelmeh
Tobias Frauenschläger
Kareem
Paul Adelsbach
David Garske
Jeremiah Mackey
Mattia Moffa
Reda Chouk
Brett Nicholas
Daniel Pouzzner
night1rider