CVE-2026-92786

LightGBM through 4.7.0 fails to validate child and split array values when parsing text models, allowing attackers to write out-of-bounds memory during SHAP prediction. Attackers can craft malicious model files with invalid node references that trigger out-of-bounds writes at attacker-chosen offsets in the leaf_depth_ buffer during feature contribution computation.

• https://github.com/lightgbm-org/LightGBM
• https://github.com/lightgbm-org/LightGBM/blob/v4.6.0/include/LightGBM/tree.h#L691-L698
• https://github.com/lightgbm-org/LightGBM/blob/v4.6.0/src/io/tree.cpp#L750-L766
• https://github.com/lightgbm-org/LightGBM/issues/7357
• https://www.vulncheck.com/advisories/lightgbm-through-4.7.0-out-of-bounds-write-via-crafted-model