Skip to content

Fix LinearRegressor heap-buffer-overflow from undersized coefficients #27964

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
bmehta001 wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Fix LinearRegressor heap-buffer-overflow from undersized coefficients #27964

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f2b60a1
Validate coefficients size in LinearRegressor to prevent OOB read
bmehta001 Apr 2, 2026
494c070
Use non-throwing overflow check for coefficients validation
bmehta001 Apr 4, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 26 additions & 0 deletions onnxruntime/core/providers/cpu/ml/linearregressor.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,8 @@
#include "core/common/narrow.h"
#include "core/providers/cpu/math/gemm.h"

#include <limits>

Check warning on line 8 in onnxruntime/core/providers/cpu/ml/linearregressor.cc

View workflow job for this annotation

GitHub Actions / Optional Lint C++ 

[cpplint] reported by reviewdog 🐶
 Found C++ system header after other header. Should be: linearregressor.h, c system, c++ system, other.  [build/include_order] [4]

Raw Output:
onnxruntime/core/providers/cpu/ml/linearregressor.cc:8:  Found C++ system header after other header. Should be: linearregressor.h, c system, c++ system, other.  [build/include_order] [4]

namespace onnxruntime {
namespace ml {

Expand Down Expand Up @@ -86,6 +88,30 @@
ptrdiff_t num_batches = input_shape.NumDimensions() <= 1 ? 1 : narrow<ptrdiff_t>(input_shape[0]);
ptrdiff_t num_features = input_shape.NumDimensions() <= 1 ? narrow<ptrdiff_t>(input_shape.Size())
: narrow<ptrdiff_t>(input_shape[1]);

// Coefficients are treated as a [num_targets, num_features] matrix.
// Validate size to prevent out-of-bounds reads in the GEMM backend.
if (num_targets_ < 0) {
return ORT_MAKE_STATUS(ONNXRUNTIME, INVALID_ARGUMENT,
"LinearRegressor: targets attribute must be non-negative, got ", num_targets_);
}

const auto targets = static_cast<size_t>(num_targets_);
const auto features = static_cast<size_t>(num_features);
// Non-throwing overflow check: targets * features would overflow if features > max / targets.
if (targets != 0 && features > std::numeric_limits<size_t>::max() / targets) {
return ORT_MAKE_STATUS(ONNXRUNTIME, INVALID_ARGUMENT,
"LinearRegressor: coefficients size overflow for targets (", num_targets_,
") * input features (", num_features, ")");
}

if (coefficients_.size() != targets * features) {
return ORT_MAKE_STATUS(ONNXRUNTIME, INVALID_ARGUMENT,
"LinearRegressor: coefficients attribute size (", coefficients_.size(),
") does not match targets (", num_targets_,
") * input features (", num_features, ")");
}
Comment thread
bmehta001 marked this conversation as resolved.

Tensor& Y = *ctx->Output(0, {num_batches, num_targets_});
concurrency::ThreadPool* tp = ctx->GetOperatorThreadPool();

Expand Down
13 changes: 13 additions & 0 deletions onnxruntime/test/providers/cpu/ml/linearregressor_test.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -85,5 +85,18 @@ INSTANTIATE_TEST_SUITE_P(
LinearRegressorParam("SOFTMAX_ZERO", {3.442477e-14f, 1.f, 1.670142e-05f, 1.f, 1.0f, 0.f}, 2)

));

// Regression test: coefficients size must match targets * num_features.
// A mismatch previously caused an out-of-bounds read in MLAS SGEMM packing.
TEST(LinearRegressorTest, CoefficientsSizeMismatch) {
OpTester test("LinearRegressor", 1, onnxruntime::kMLDomain);
// 1 coefficient but input has 2 features and targets=1 → expects 2 coefficients
test.AddAttribute("coefficients", std::vector<float>{1.0f});
test.AddAttribute("targets", int64_t{1});
test.AddInput<float>("X", {1, 2}, {1.f, 2.f});
test.AddOutput<float>("Y", {1, 1}, {0.f});
test.Run(OpTester::ExpectResult::kExpectFailure, "coefficients attribute size");
}

} // namespace test
} // namespace onnxruntime
Loading