Affected versions of this crate tried to preallocate a vector for an arbitrary amount of bytes announced by the ASN.1-DER length field without further checks.

This allows an attacker to trigger a SIGABRT by creating length fields that announce more bytes than the allocator can provide.

The flaw was corrected by not preallocating memory.

A bug in the SmallVec::insert_many method caused it to allocate a buffer that was smaller than needed. It then wrote past the end of the buffer, causing a buffer overflow and memory corruption on the heap.

This bug was only triggered if the iterator passed to insert_many yielded more items than the lower bound returned from its size_hint method.

The flaw was corrected in smallvec 0.6.14 and 1.6.1, by ensuring that additional space is always reserved for each item inserted. The fix also simplified the implementation of insert_many to use less unsafe code, so it is easier to verify its correctness.

Thank you to Yechan Bae (@Qwaz) and the Rust group at Georgia Tech’s SSLab for finding and reporting this bug.

If a tokio::sync::oneshot channel is closed (via the oneshot::Receiver::close method), a data race may occur if the oneshot::Sender::send method is called while the corresponding oneshot::Receiver is awaited or calling try_recv.

When these methods are called concurrently on a closed channel, the two halves of the channel can concurrently access a shared memory location, resulting in a data race. This has been observed to cause memory corruption.

Note that the race only occurs when both halves of the channel are used after the Receiver half has called close. Code where close is not used, or where the Receiver is not awaited and try_recv is not called after calling close, is not affected.

See tokio#4225 for more details.

Impact

Untrusted websocket connections can cause an out-of-memory (OOM) process abort in a client or a server. The root cause of the issue is during dataframe parsing. Affected versions would allocate a buffer based on the declared dataframe size, which may come from an untrusted source. When Vec::with_capacity fails to allocate, the default Rust allocator will abort the current process, killing all threads. This affects only sync (non-Tokio) implementation. Async version also does not limit memory, but does not use with_capacity, so DoS can happen only when bytes for oversized dataframe or message actually got delivered by the attacker.

This is a security concern for you, if

• your server application handles untrusted websocket connections
• OR your client application connects to untrusted websocket servers

Patches

The crashes are fixed in version 0.26.5 by imposing default dataframe size limits. Affected users are advised to update to this version.

Note that default memory limits are rather large (100MB dataframes and 200 MB messages), so they can still cause DoS in some environments (i.e. 32-bit). New API has been added to fine tune those limits for specific applications.

Workarounds

• Migrate your project to another, maintained Websocket library like Tungstenite.
• Accept only trusted WebSocket traffic.
• Filter the WebSocket traffic though some kind of proxy that ensures sanity limits on messages.
• Handle process aborts gracefully and limit process memory using OS tools.

Credits

This issue was reported by Evan Richter at ForAllSecure and found with Mayhem and Cargo Fuzz.

Versions of ed25519-dalek prior to v2.0 model private and public keys as separate types which can be assembled into a Keypair, and also provide APIs for serializing and deserializing 64-byte private/public keypairs.

Such APIs and serializations are inherently unsafe as the public key is one of the inputs used in the deterministic computation of the S part of the signature, but not in the R value. An adversary could somehow use the signing function as an oracle that allows arbitrary public keys as input can obtain two signatures for the same message sharing the same R and only differ on the S part.

Unfortunately, when this happens, one can easily extract the private key.

Revised public APIs in v2.0 of ed25519-dalek do NOT allow a decoupled private/public keypair as signing input, except as part of specially labeled "hazmat" APIs which are clearly labeled as being dangerous if misused.

There was a logic bug where unauthenticated payloads could still cause a nonce increment in snow's internal state. For an attacker with privileges to inject packets into the channel over which the Noise session operates, this could allow a denial-of-service attack which could prevent message delivery by sending garbage data.

Note that this only affects those who are using the stateful TransportState, not those using StatelessTransportState.

This has been patched in version 0.9.5, and all users are recommended to update.

Timing variability of any kind is problematic when working with potentially secret values such as elliptic curve scalars, and such issues can potentially leak private keys and other secrets. Such a problem was recently discovered in curve25519-dalek.

The Scalar29::sub (32-bit) and Scalar52::sub (64-bit) functions contained usage of a mask value inside a loop where LLVM saw an opportunity to insert a branch instruction (jns on x86) to conditionally bypass this code section when the mask value is set to zero as can be seen in godbolt:

• 32-bit (see L106): https://godbolt.org/z/zvaWxzvqv
• 64-bit (see L48): https://godbolt.org/z/PczYj7Pda

A similar problem was recently discovered in the Kyber reference implementation:

https://groups.google.com/a/list.nist.gov/g/pqc-forum/c/hqbtIGFKIpU/m/cnE3pbueBgAJ

As discussed on that thread, one portable solution, which is also used in this PR, is to introduce a volatile read as an optimization barrier, which prevents the compiler from optimizing it away.

The fix can be validated in godbolt here:

• 32-bit: https://godbolt.org/z/jc9j7eb8E
• 64-bit: https://godbolt.org/z/x8d46Yfah

The problem was discovered and the solution independently verified by Alexander Wagner [email protected] and Lea Themint [email protected] using their DATA tool:

https://github.com/Fraunhofer-AISEC/DATA

Affected version of this crate did not properly parse unknown fields when parsing a user-supplied input.

This allows an attacker to cause a stack overflow when parsing the mssage on untrusted data.

ring::aead::quic::HeaderProtectionKey::new_mask() may panic when overflow checking is enabled. In the QUIC protocol, an attacker can induce this panic by sending a specially-crafted packet. Even unintentionally it is likely to occur in 1 out of every 2**32 packets sent and/or received.

On 64-bit targets operations using ring::aead::{AES_128_GCM, AES_256_GCM} may panic when overflow checking is enabled, when encrypting/decrypting approximately 68,719,476,700 bytes (about 64 gigabytes) of data in a single chunk. Protocols like TLS and SSH are not affected by this because those protocols break large amounts of data into small chunks. Similarly, most applications will not attempt to encrypt/decrypt 64GB of data in one chunk.

Overflow checking is not enabled in release mode by default, but RUSTFLAGS="-C overflow-checks" or overflow-checks = true in the Cargo.toml profile can override this. Overflow checking is usually enabled by default in debug mode.