Python (and maybe later Rust, maybe even Go) implementation of Shamir Secret Sharing, as specified in the KMIP v2.0 protocol
Plans:
- Add higher level API (i.e. split/recover for full value, not just slice)
- Add test suite (started with testdocs)
- Write user docs
- PyPi (?)
- Rewrite in a more conventional language (Rust maybe? Probably not C. Not COBOL, so don't ask)
Progress:
- Rewrite and tighten up shamirshare.py as shamirshare2.py (done)
- Initial test suite deployed (more work needed)
The original implementation (shamirshare.py) works, but was utterly Baroque (good for music, less so for code). The rewrite (shamirshare2.py) removes functions not needed for Shamir sharing, removes most operator overloading and as much coercion code as possible - all things which will be hard to port to more conventional languages and are prone to bugs in any language.
Current status: ver 0.33, 8 Oct 2019:
- More minor edits to make the comments doctest friendly
ver 0.32, 27 Sept 2019:
- Minor edits to make the comments doctest friendly
ver 0.31, 25 Sept 2019:
- Replaced home-rolled Python2/3 compatible functions with six package
ver 0.3, 22 Sept 2019:
- Implemented GF16
ver 0.2, 21 Sept 2019:
- Implemented GF8
ver 0.1, 18 Sept 2019:
- Implemented GFp, fit and eval functions
- Need to implement GF8 and GF16
Current status: ver 0.3, 11 Sept 2019:
- Implemented GF16
- Essentially done with this verbose version (time to implement in simpler Python, then ...)
ver 0.21, 7 Sept 2019:
- Basic GFp and GF8 functionality working (need to implement GF16)
- Python 2 Singleton issue fixed
...
ver 0.1, 18 Aug 2019:
- Initial commit
- Implemented GF8 base field, fit and call (aka eval) functions
Normally your secret will be larger than a single 8-bit byte. Divide the secret into byte-size slices, and split each byte separately over GF8. For each user the split bytes can themselves be combined to form a larger split. The same approach can be used for GF16 - splitting the secret into 16-bit slices.
Example - GF8 3-way (i.e. quadratic) split of 32-bit secret (i.e. 4 slices)
>>> thesplits = [[[gf8(user), gf8(randint(0,255))] for user in range(1,4)] for slice in range(4)]
>>> [[[format(thesplits[slice][user][0]),format(thesplits[slice][user][1])] for user in range(3)] for slice in range(4)]
[
[['01', 'b2'], ['02', 'd7'], ['03', 'e8']],
[['01', 'f0'], ['02', '0e'], ['03', '9e']],
[['01', 'ea'], ['02', '7d'], ['03', '28']],
[['01', '76'], ['02', 'b8'], ['03', 'b8']]
]
User #1 has array of secrets ['b2','f0','ea','76'], user #2 has ['d7','0e','7d','b8'], etc.
>>> pfit0 = fit(thesplits[0]); list(map(format,pfit0))
['8d', '31', '0e']
So first byte (slice) is split by polynomial 8d + 31*x + 0e*x^2, and has secret value '8d'. The total 32-bit secret is the array ['8d','60','bf','76'].
>>> pfit = [fit(thesplits[slice]) for slice in range(4)]
>>> list(map(format,[pfit[slice][0] for slice in range(4)]))
['8d', '60', 'bf', '76']
As usual, a pre-existing secret can be split by making one of the splits for a “user #0”.