[WIP]: switch to crypto-bigint for decryption

[dignifiedquire]

Very, very WIP

Uncomplete, unordered task list

• switch internal storage for RsaPrivateKey
• switch internal storage for RsaPublicKey
• switch all code to use the new decrypt implementation
• update public traits using BigUint to return owned versions
• fix blinding implementation
• switch decryption algorithm with precompute to use crypto-bigint ops
• go through other algorithms and update what can be done without having primality checks implemented
• review & update code for constant time operation
• review & update code for performance
• benchmarks

Tests

• algorithms::pad::tests::test_left_pad
• algorithms::pkcs1v15::tests::test_non_zero_bytes
• algorithms::rsa::tests::recover_primes_works
• key::tests::build_key_from_p_q
• key::tests::build_key_from_primes
• key::tests::invalid_coeff_private_key_regression
• algorithms::generate::tests::key_generation_128
• key::tests::key_generation_128
• algorithms::generate::tests::test_impossible_keys
• algorithms::generate::tests::key_generation_multi_3_256
• algorithms::generate::tests::key_generation_multi_4_64
• key::tests::key_generation_multi_4_64
• key::tests::reject_oversized_private_key
• key::tests::test_from_into
• key::tests::key_generation_multi_3_256
• key::tests::test_serde
• oaep::decrypting_key::tests::test_serde
• oaep::encrypting_key::tests::test_serde
• oaep::tests::test_decrypt_oaep_invalid_hash
• oaep::tests::test_decrypt_oaep_invalid_hash_traits
• oaep::tests::test_encrypt_decrypt_oaep
• oaep::tests::test_encrypt_decrypt_oaep_traits
• pkcs1v15::decrypting_key::tests::test_serde
• pkcs1v15::encrypting_key::tests::test_serde
• pkcs1v15::signature::tests::test_serde
• pkcs1v15::signing_key::tests::test_serde
• pkcs1v15::tests::test_decrypt_pkcs1v15
• pkcs1v15::tests::test_decrypt_pkcs1v15_traits
• pkcs1v15::tests::test_encrypt_decrypt_pkcs1v15
• pkcs1v15::tests::test_encrypt_decrypt_pkcs1v15_traits
• pkcs1v15::tests::test_sign_pkcs1v15
• pkcs1v15::tests::test_sign_pkcs1v15_digest_signer
• pkcs1v15::tests::test_sign_pkcs1v15_signer
• pkcs1v15::tests::test_sign_pkcs1v15_signer_sha2_256
• pkcs1v15::tests::test_sign_pkcs1v15_signer_sha3_256
• pkcs1v15::tests::test_unpadded_signature
• pkcs1v15::tests::test_unpadded_signature_hazmat
• pkcs1v15::tests::test_verify_pkcs1v15
• pkcs1v15::tests::test_verify_pkcs1v15_digest_signer
• pkcs1v15::tests::test_verify_pkcs1v15_signer
• pkcs1v15::verifying_key::tests::test_serde
• pss::blinded_signing_key::tests::test_serde
• pss::signature::tests::test_serde
• pss::signing_key::tests::test_serde
• algorithms::generate::tests::key_generation_1024
• pss::test::test_sign_and_verify_pss_blinded_hazmat
• pss::test::test_sign_and_verify_pss_hazmat
• pss::test::test_sign_and_verify_roundtrip
• pss::test::test_sign_and_verify_roundtrip_blinded_digest_signer
• pss::test::test_sign_and_verify_roundtrip_blinded_signer
• pss::test::test_sign_and_verify_roundtrip_digest_signer
• pss::test::test_sign_and_verify_roundtrip_signer
• pss::test::test_sign_blinded_and_verify_roundtrip
• pss::test::test_verify_pss
• pss::test::test_verify_pss_digest_signer
• pss::test::test_verify_pss_hazmat
• pss::test::test_verify_pss_signer
• algorithms::generate::tests::key_generation_multi_5_64
• pss::verifying_key::tests::test_serde
• key::tests::test_negative_decryption_value
• key::tests::key_generation_multi_5_64
• pss::test::test_sign_and_verify_2049bit_key
• key::tests::key_generation_1024
• algorithms::generate::tests::key_generation_multi_8_576
• key::tests::key_generation_multi_8_576

[tarcieri]

@dignifiedquire I think it'd be good to land a spike / PoC of this even if all the algorithms aren't yet implemented, since it would unblock splitting up the work on (re)implementing those

[gogo2464]

@tarcieri can I be assigned to the poc please? I would like to gain experience.

EDIT: I did not undertand the message yet. If the exploit is already made, yes I can read it. Just do not reinvent the whell. I was already working on.

EDIT2: Can you just provide me as much tools as possible please? I am currenyly reading the doc of marvin-tool

[gogo2464]

I can see difference but I would like a tool or a procedure to analyze the time spent. I am currently working with callgrind and kcachegrind.

[dignifiedquire]

@tarcieri most things are working now 🎉

I would appreciate some help with debugging the last failures, seems the proptests are discovering some roundtrip issues in the encoding/decoding and the last regular test that is a problem is dealign with a 2049bit key

[dignifiedquire]

and also feel free to start to review the code/fix up anything you think needs improving

[kjvalencik]

This is awesome! I'm excited to see this PR progressing so much.

seems the proptests are discovering some roundtrip issues in the encoding/decoding and the last regular test that is a problem is dealign with a 2049bit key

I looked into this for a few minutes and these two appear to be different cases of the same problem. crypto-bigint uses 64-bit limbs, which requires it to round higher on number of bytes. For example, with a 2049-bit key it's 264 instead of 257; leading to more padding in the serialized version. This may be the same thing causing round-trip problems.

[Fethbita]

I don't think it's good to limit e to 64 bits. See RFC 8017 Section 3.1:

In a valid RSA public key, the RSA modulus n is a product of u distinct odd primes r_i, i = 1, 2, ..., u, where u >= 2, and the RSA public exponent e is an integer between 3 and n - 1 satisfying GCD(e,\lambda(n)) = 1, where \lambda(n) = LCM(r_1 - 1, ..., r_u - 1). By convention, the first two primes r_1 and r_2 may also be denoted p and q, respectively.

And FIPS 186-4 page 52

The exponent e shall be an odd positive integer such that:
2¹⁶ < e < 2²⁵⁶.

It should still be BoxedUint.

[dignifiedquire]

thanks for pointing this out, our default safe constants restricting it, but you are right, it should be possible for larger exponents. this is now fixed

[zeerooth]

I tried testing this PR in a no_std environment, but it doesn't compile, as RSA pulls getrandom via crypto-bigint even if I disable default features.

rsa = { git = "https://github.com/RustCrypto/RSA.git", branch = "const-crypto-biguint", default-features = false }

    rsa v0.10.0-pre.2 (https://github.com/RustCrypto/RSA.git?branch=const-crypto-biguint#74d31971)
    ├── const-oid v0.10.0-rc.0
    ├── crypto-bigint v0.6.0-rc.5
    │   ├── num-traits v0.2.19
    │   │   [build-dependencies]
    │   │   └── autocfg v1.4.0
    │   ├── rand_core v0.6.4
    │   │   └── getrandom v0.2.15
    │   │       └── cfg-if v1.0.0
    │   ├── subtle v2.6.1
    │   └── zeroize v1.8.1

error: target is not supported, for more information see: https://docs.rs/getrandom/#unsupported-targets
   --> /home/----/.cargo/registry/src/index.crates.io-6f17d22bba15001f/getrandom-0.2.15/src/lib.rs:347:9
    |
347 | /         compile_error!("target is not supported, for more information see: \
348 | |                         https://docs.rs/getrandom/#unsupported-targets");
    | |________________________________________________________________________^

[baloo]

Float::ln was used to provide support for f64::ln on non-std (used line 45).

[mepi262]

@dignifiedquire
Any update on this pull request?

[dignifiedquire]

@dignifiedquire Any update on this pull request?

no, haven't had time to fix the afformentioned issues yet

[mepi262]

@dignifiedquire
Thank you for your response !
I hope your another project will be succeed and become be able to concentrate on this pull request.

[dignifiedquire]

@tarcieri current benchmarks

# crypto-bigint

# macbook m1
test bench_rsa_2048_pkcsv1_decrypt      ... bench:   7,184,387.50 ns/iter (+/- 425,598.69)
test bench_rsa_2048_pkcsv1_sign_blinded ... bench:  13,453,579.10 ns/iter (+/- 686,276.31)

# AMD
test bench_rsa_2048_pkcsv1_decrypt      ... bench:   9,260,832.80 ns/iter (+/- 30,013.38)
test bench_rsa_2048_pkcsv1_sign_blinded ... bench:  16,610,079.40 ns/iter (+/- 251,292.53)

# master

# macbook m1
test bench_rsa_2048_pkcsv1_decrypt      ... bench:   1,117,479.15 ns/iter (+/- 31,334.30)
test bench_rsa_2048_pkcsv1_sign_blinded ... bench:   1,337,437.55 ns/iter (+/- 88,624.39)

# AMD
test bench_rsa_2048_pkcsv1_decrypt      ... bench:   1,414,348.80 ns/iter (+/- 12,585.71)
test bench_rsa_2048_pkcsv1_sign_blinded ... bench:   1,685,650.00 ns/iter (+/- 11,105.71)