CVE-2026-55093

Tract is a tiny, no-nonsense, self-contained TensorFlow and ONNX inference toolkit. Prior to 0.21.16, 0.22.2, and 0.23.1, tract-nnef uses unchecked usize multiplication in nnef/src/tensors.rs read_tensor for attacker-controlled tensor dimensions, the allocation size, and the reported tensor length. Loading a crafted NNEF archive through model_for_path or model_for_read reaches the default DatLoader and can make the wrapped size check accept a small allocation while data/src/tensor.rs as_slice_unchecked creates a much larger logical slice. Model construction through as_uniform can then read beyond the heap allocation and disclose adjacent data, and later access can terminate the process with a segmentation fault. The affected dense numeric tensor path does not include the independently guarded bool, String, or block-quant paths, and no out-of-bounds write or code execution was demonstrated. This issue is fixed in versions 0.21.16, 0.22.2, and 0.23.1.

Tract is a tiny, no-nonsense, self-contained TensorFlow and ONNX inference toolkit. Prior to 0.21.16, 0.22.2, and 0.23.1, tract-nnef uses unchecked usize multiplication in nnef/src/tensors.rs read_tensor for attacker-controlled tensor dimensions, the allocation size, and the reported tensor length. Loading a crafted NNEF archive through model_for_path or model_for_read reaches the default DatLoader and can make the wrapped size check accept a small allocation while data/src/tensor.rs as_slice_unchecked creates a much larger logical slice. Model construction through as_uniform can then read beyond the heap allocation and disclose adjacent data, and later access can terminate the process with a segmentation fault. The affected dense numeric tensor path does not include the independently guarded bool, String, or block-quant paths, and no out-of-bounds write or code execution was demonstrated. This issue is fixed in versions 0.21.16, 0.22.2, and 0.23.1.

• Component: tract-nnef (nnef/src/tensors.rs::read_tensor) + tract-data (data/src/tensor.rs)
• Affected versions: < 0.21.16, 0.22.0–0.22.2, 0.23.0–0.23.1 — the dense DatLoader path was unguarded across all three release lines; patched in 0.21.16 / 0.22.2 / 0.23.1
• Class: CWE-190 (integer overflow) → CWE-125 (out-of-bounds read)
• Trigger: loading a crafted NNEF model archive (*.nnef.tgz / *.nnef.tar / dir) via the public tract_nnef::nnef().model_for_path / model_for_read
• Impact: read_tensor returns a memory-unsafe tensor (reported len 2^61 over a 56-byte heap allocation). Always-on primitive: a bounded heap out-of-bounds read during model build (as_uniform), an adjacent-heap information-disclosure reachable via the public load API. The resulting slice is an unsound from_raw_parts(ptr, 2^61) that SIGSEGVs (DoS) on any access past the mapped region (demonstrated by direct access). No out-of-bounds write and no RCE were achieved — tract's const-folding/as_uniform fast-paths fold simple consuming graphs without the full read.
• Severity: Medium

Summary

read_tensor builds a tensor shape from attacker-controlled 32-bit dimensions and computes the element count len = product(shape) and the byte allocation product(shape) * size_of(dt) with unchecked usize arithmetic. In --release (no overflow-checks), both products wrap modulo 2^64. An attacker chooses dimensions so that the wrapped products collapse to a small value that satisfies the header consistency check, while the true element count remains astronomically large. read_tensor returns Ok with a Tensor whose reported len (e.g. 2^61+7) is far larger than its backing heap allocation (e.g. 56 bytes). The unchecked slice accessor as_slice_unchecked (from_raw_parts(ptr, self.len)) then produces a slice spanning ~18 exabytes over a 56-byte buffer. The out-of-bounds read fires automatically during model build (no inference required), reachable through the default DatLoader resource loader.

Root cause

nnef/src/tensors.rs, read_tensor:

let shape: TVec<usize> = header.dims[0..header.rank as usize].iter().map(|d| *d as _).collect();
let len = shape.iter().product::<usize>();                       // (1) unchecked, wraps
...
} else if header.bits_per_item != u32::MAX
    && len * (header.bits_per_item as usize / 8) != header.data_size_bytes as usize  // (2) wrapped == u32
{
    bail!(...);
}
...
let mut tensor = unsafe { Tensor::uninitialized_dt(dt, &shape)? };   // (3) alloc off the same wrapped product
...
reader.read_exact(plain.as_bytes_mut())?;                            // storage-bounded read, no overflow here
Ok(tensor)

data/src/tensor.rs, uninitialized_aligned_dt:

let bytes = shape.iter().cloned().product::<usize>() * dt.size_of();  // (3) wraps to the same small value
let storage = ... Blob::new_for_size_and_align(bytes, alignment) ...;
...
tensor.update_strides_and_len();                                     // len = product(shape), wraps, no clamp

The three quantities — the consistency-check LHS (2), the allocation (3), and the reported len — are all the same wrapped product(shape)*size_of, so they stay mutually consistent and the consistency check at (2) cannot catch the overflow. data_size_bytes is a u32, so the attacker simply sets it to the wrapped value.

Corruption sink — data/src/tensor.rs::as_slice_unchecked (and data/src/tensor/plain_view.rs::as_slice_unchecked):

if self.storage.byte_len() == 0 { &[] }
else { std::slice::from_raw_parts(self.as_ptr_unchecked(), self.len()) }  // len = 2^61 over a 56-byte alloc

The only guard is byte_len() == 0. A small non-zero allocation defeats it and yields an unsound oversized slice.

Witness (F64)

dims          = [33955849, 7005787, 359, 3, 3, 3]   (rank 6, each <= u32::MAX)
product(shape)= 2_305_843_009_213_693_959 = 2^61 + 7…

Tract is a tiny, no-nonsense, self-contained TensorFlow and ONNX inference toolkit. Prior to 0.21.16, 0.22.2, and 0.23.1, tract-nnef uses unchecked usize multiplication in nnef/src/tensors.rs read_tensor for attacker-controlled tensor dimensions, the allocation size, and the reported tensor length. Loading a crafted NNEF archive through model_for_path or model_for_read reaches the default DatLoader and can make the wrapped size check accept a small allocation while data/src/tensor.rs as_slice_unchecked creates a much larger logical slice. Model construction through as_uniform can then read beyond the heap allocation and disclose adjacent data, and later access can terminate the process with a segmentation fault. The affected dense numeric tensor path does not include the independently guarded bool, String, or block-quant paths, and no out-of-bounds write or code execution was demonstrated. This issue is fixed in versions 0.21.16, 0.22.2, and 0.23.1.

Tract is a tiny, no-nonsense, self-contained TensorFlow and ONNX inference toolkit. Prior to 0.21.16, 0.22.2, and 0.23.1, tract-nnef uses unchecked usize multiplication in nnef/src/tensors.rs read_tensor for attacker-controlled tensor dimensions, the allocation size, and the reported tensor length. Loading a crafted NNEF archive through model_for_path or model_for_read reaches the default DatLoader and can make the wrapped size check accept a small allocation while data/src/tensor.rs as_slice_unchecked creates a much larger logical slice. Model construction through as_uniform can then read beyond the heap allocation and disclose adjacent data, and later access can terminate the process with a segmentation fault. The affected dense numeric tensor path does not include the independently guarded bool, String, or block-quant paths, and no out-of-bounds write or code execution was demonstrated. This issue is fixed in versions 0.21.16, 0.22.2, and 0.23.1.

• Component: tract-nnef (nnef/src/tensors.rs::read_tensor) + tract-data (data/src/tensor.rs)
• Affected versions: < 0.21.16, 0.22.0–0.22.2, 0.23.0–0.23.1 — the dense DatLoader path was unguarded across all three release lines; patched in 0.21.16 / 0.22.2 / 0.23.1
• Class: CWE-190 (integer overflow) → CWE-125 (out-of-bounds read)
• Trigger: loading a crafted NNEF model archive (*.nnef.tgz / *.nnef.tar / dir) via the public tract_nnef::nnef().model_for_path / model_for_read
• Impact: read_tensor returns a memory-unsafe tensor (reported len 2^61 over a 56-byte heap allocation). Always-on primitive: a bounded heap out-of-bounds read during model build (as_uniform), an adjacent-heap information-disclosure reachable via the public load API. The resulting slice is an unsound from_raw_parts(ptr, 2^61) that SIGSEGVs (DoS) on any access past the mapped region (demonstrated by direct access). No out-of-bounds write and no RCE were achieved — tract's const-folding/as_uniform fast-paths fold simple consuming graphs without the full read.
• Severity: Medium

Summary

read_tensor builds a tensor shape from attacker-controlled 32-bit dimensions and computes the element count len = product(shape) and the byte allocation product(shape) * size_of(dt) with unchecked usize arithmetic. In --release (no overflow-checks), both products wrap modulo 2^64. An attacker chooses dimensions so that the wrapped products collapse to a small value that satisfies the header consistency check, while the true element count remains astronomically large. read_tensor returns Ok with a Tensor whose reported len (e.g. 2^61+7) is far larger than its backing heap allocation (e.g. 56 bytes). The unchecked slice accessor as_slice_unchecked (from_raw_parts(ptr, self.len)) then produces a slice spanning ~18 exabytes over a 56-byte buffer. The out-of-bounds read fires automatically during model build (no inference required), reachable through the default DatLoader resource loader.

Root cause

nnef/src/tensors.rs, read_tensor:

let shape: TVec<usize> = header.dims[0..header.rank as usize].iter().map(|d| *d as _).collect();
let len = shape.iter().product::<usize>();                       // (1) unchecked, wraps
...
} else if header.bits_per_item != u32::MAX
    && len * (header.bits_per_item as usize / 8) != header.data_size_bytes as usize  // (2) wrapped == u32
{
    bail!(...);
}
...
let mut tensor = unsafe { Tensor::uninitialized_dt(dt, &shape)? };   // (3) alloc off the same wrapped product
...
reader.read_exact(plain.as_bytes_mut())?;                            // storage-bounded read, no overflow here
Ok(tensor)

data/src/tensor.rs, uninitialized_aligned_dt:

let bytes = shape.iter().cloned().product::<usize>() * dt.size_of();  // (3) wraps to the same small value
let storage = ... Blob::new_for_size_and_align(bytes, alignment) ...;
...
tensor.update_strides_and_len();                                     // len = product(shape), wraps, no clamp

The three quantities — the consistency-check LHS (2), the allocation (3), and the reported len — are all the same wrapped product(shape)*size_of, so they stay mutually consistent and the consistency check at (2) cannot catch the overflow. data_size_bytes is a u32, so the attacker simply sets it to the wrapped value.

Corruption sink — data/src/tensor.rs::as_slice_unchecked (and data/src/tensor/plain_view.rs::as_slice_unchecked):

if self.storage.byte_len() == 0 { &[] }
else { std::slice::from_raw_parts(self.as_ptr_unchecked(), self.len()) }  // len = 2^61 over a 56-byte alloc

The only guard is byte_len() == 0. A small non-zero allocation defeats it and yields an unsound oversized slice.

Witness (F64)

dims          = [33955849, 7005787, 359, 3, 3, 3]   (rank 6, each <= u32::MAX)
product(shape)= 2_305_843_009_213_693_959 = 2^61 + 7…

tract_nnef::tensors::read_tensor builds a tensor shape from attacker-controlled 32-bit dimensions and computes both the element count product(shape) and the byte allocation product(shape) * size_of(dt) with unchecked usize arithmetic. In release builds (no overflow-checks) both products wrap modulo 2^64.

A crafted NNEF .dat tensor can choose dimensions whose wrapped products collapse to a small value that satisfies the header size-consistency check, while the true element count stays astronomically large. read_tensor then returns a Tensor whose reported len (e.g. 2^61 + 7) far exceeds its backing heap allocation (e.g. 56 bytes). The unchecked accessor as_slice_unchecked (slice::from_raw_parts(ptr, self.len())) subsequently yields a slice spanning ~18 EiB over the small buffer.

The out-of-bounds read fires automatically during model build (no inference required), reachable through the default DatLoader resource loader via the public tract_nnef::nnef().model_for_path / model_for_read API when the const-folding as_uniform fast-path materializes the over-long constant. The always-on primitive is a bounded adjacent-heap over-read (information disclosure); access further past the mapped region SIGSEGVs (denial of service). No out-of-bounds write or code execution was demonstrated.

Affected: every release line prior to the backported fixes — < 0.21.16, 0.22.0–0.22.1, and 0.23.0. The block-quant path had already received an analogous blob-size guard; the dense DatLoader path was missed.

Mitigation

Upgrade to 0.21.16, 0.22.2, or 0.23.1. The fix computes the shape product and byte size with checked_mul and rejects the tensor on overflow (commit 34c7df2).