This is useful for programs/protocol which can do some heuristics or
learning-based approach towards optimising the table size based on the number of
uses of e.g. a public key, the second basepoint in a Pedersen commitment, etc.,
i.e. the first time a public key is used to verify a signature, the usual
variable-time basepoint multiscalar multiplication is used, however after 1000
verifications, the table size is upgraded, and again after 10000 verifications,
etc.
This implements a macro for implementing the BasepointTable trait, and
uses the macro to create basepoint table types. The default table
still uses radix-16 representation and is ~30KB in size. The new
table types, and their memory usage and additions required per
basepoint multiplication are:
* `EdwardsBasepointTableRadix64`: ~120KB, 43 additions
* `EdwardsBasepointTableRadix128`: ~240KB, 37 additions
* `EdwardsBasepointTableRadix256`: ~480KB, 32 additions
The functionality we were using is now contained in the `rand_core` crate, which
we already depend upon. As far as testing code goes, only benchmarks still
depend upon `rand`, as they use `thread_rng`.
We due this in lieu of implementing `TryFrom` to allow for API
consumers to use the `?` operator to convert potential `None`s into
their own `Result<T, CustomError>` types for better error handling
with less boilerplate.
Note that this is a breaking API change.
This test is no longer necessary as it originally had a possibility of
failure due to the original implementation of the invsqrt() function
in the decompression algorithm. The failure with the current API is
nonexistent.
* FIXES#288.
`zeroize` is WASM-friendly as it has no dependencies on C compilers.
Instead uses Rust's own volatile write semantics and compiler fences to
ensure zeroization is not elided by the compiler.
This is a breaking change to the serialization format. It fixes it so that the
Serde encoding can match the conventional encoding of each type of object, and
so that Serde can be used with no overhead -- when using serde-bincode, the
Serde encoding now matches the manual encoding.
This ensures that the serde Serialize and Deserialize implementations use
fixed-length Serde tuples, rather than variable-length byte arrays. This flaw
in data modeling was pointed out by Trevor Perrin.
This was more useful at the time when we were determining, e.g., optimal lookup
table sizes and could regenerate them more easily, but it came at a massive
complexity cost. It also meant that we were unable to implement backend
autoselection. This commit removes the `build.rs` entirely. In the future, a
different `build.rs` could be added that auto-selects a backend, but it seems
like the current default-u64 setup has been working fine.
Closes#238.
This issue was discovered independently by both Jack "str4d" Grigg
(issue #238), who noted that reduction was not performed on addition, and
Laurent Grémy & Nicolas Surbayrole of Quarkslab, who noted that it was possible
to cause an overflow and compute incorrect results.
Found by @3for; this only affected width-7 NAF computations, which were never
used in the source tree (only width 5, optimal for dynamic cases, and 8, better
for static cases).
Closes#272
The function `FieldElement::sqrt_ratio_i` always returns a positive root
by definition. Therefore the test for negativity in the edwards point
decompression function always returns false and we only need to flip its
sign if `compressed_sign_bit` is set.
The NAF computation can generate a 1 in the last digit (only) when s = 2^255-1,
so someone who manually constructed the value s = 2^255-1 and fed it into a NAF-using
computation could generate an incorrect result. Some version of this bug has
been present from the beginning of the library, but it has no security content,
because the NAF computations are not applied to secret data, and the error
occurs only on one value which is not constructed by any client caller.
When using Scalar::from_bits to manually create unreduced Scalars (e.g.,
X/Ed25519 keys with specified bit patterns), it's possible to construct Scalar
values that range up to 2^255-1. These shouldn't ever end up in a vartime
multiscalar mul call anyways, because it doesn't handle secret data, but it is
technically allowed by the type system and should be handled. When w=8, these
can generate terminal carries that can't be folded into the last digit, but
this can be handled by folding them into an extra digit instead.
This fixes a bug in the Pippenger implementation reported by Fernando Krell and
diagnosed by Oleg Andreev. The problem is that at the largest problem sizes
(using w=8), the signed digits fill the value range of an i8, and so doing
computation on them to calculate the bucket index can hit an overflow.
This was not caught in CI because the test suite didn't check all problem
sizes; tests for these sizes which expose this bug were added in the previous
commit.
This saves 4 point doublings by unwrapping the first loop iteration,
as well as 63 field multiplications (one per iteration) by managing
curve model choice explicitly.
See discussion at https://github.com/dalek-cryptography/curve25519-dalek/issues/232 , copied below:
`1.1` changed the trait bounds for `RistrettoPoint::random` and `Scalar::random`, see #222 and #219.
These changes have two benefits:
* they unlink us from the `rand` crate and make us depend only on `rand_core`;
* they allow passing both owned and borrowed RNGs.
The change was not supposed to be a breaking change, since the new bounds are strictly more general than the old ones (as every `RngCore` is an `Rng` and every `&mut RngCore` is an `RngCore`), so the new bound is satisfied in every situation where the old bound applied.
The `1.1.0-pre.0` version didn't cause problems on the crates I tested it on, but there was an unexpected problem: ce71c93a9a/spacesuit/src/value.rs (L160-L161) broke, since it took a borrow as input and used it twice. So there was slight breakage.
One option is to revert the changes (probably just the ones from #219) and release 1.1.3; another would be to fix up `slingshot` and leave the new bound.
Building the docs currently doesn't work, because rustdoc enables parts of the
code (to document them) which then don't check (because there are missing
exports or dependencies). This **should** fix the issue, but there's no way to
test without publishing a new version.
This doesn't (yet) give any speedup over the non-precomputed multiscalar
multiplication, and it's not clear that it's a good idea to commit to
supporting it in the future. Removing it means that it's not committed-to as
part of the public API, but the source is still there in the tree if we want to
revisit it later.
Since computation of the 0 term in reduction requires a multiplication with a
4-cycle latency, this ensures that the rest of the computation can start before
the 0 term is finished
This splits the `FieldElement51x4` type into two types:
- `F51x4Reduced` (with reduced limbs)
- `F51x4Unreduced` (with unreduced limbs)
The reduction is implemented as a `From` impl to convert one type to the other.
The output of a multiplication is now a `F51x4Unreduced`. The reason is that
the inputs to IFMA operations must be at most 52 bits, so it's not possible to
perform an addition of (51+epsilon)-bit values and still be small enough to be
used as an input to multiplication. So, it doesn't make sense to perform a
reduction at the end of a multiplication, because the reduced values will be
fed into an addition or subtraction, which then needs to be re-reduced.
This begins to attempt to restructure the source tree so that the common parts
are common and the different parts are different.
The backend is now split into two parts:
- serial (containing the implementation using serial formulas and mixed-model arithmetic).
- vector (containing the implementation using parallel formulas and single-model arithmetic).
The serial scalar_mul tree is now under backend::serial::scalar_mul.
The avx2 scalar_mul tree is now under backend::avx2::scalar_mul.
`FieldElement32` -> `FieldElement2625`
`FieldElement64` -> `FieldElement51`
`Scalar32` -> `Scalar29`
`Scalar64` -> `Scalar52`
This naming is more accurate and would let us add an ADX backend later.
Vicariously updates to `generic-array` 0.12, however this change also
removes `generic-array` as a direct dependency, as it can be sourced
from the `digest` crate.
