[tensorflow/tensorflow] Integer overflow in Detection PostProcess max_detections * max_classes_per_detection causes heap buffer overflow

### ROOT CAUSE The issue arises when the product of `max_detections` and `max_classes_per_detection` exceeds the maximum value for a `size_t` (typically 2^64-1). This overflow can lead to a negative or incorrect value, causing a heap buffer overflow during memory allocation. The root cause is the unsafe multiplication without checking for integer overflow, which can be triggered by large values of `max_detections` and `max_classes_per_detection`. ### CODE FIX ```cpp #include #include // Replace the unsafe multiplication with a safe check size_t num_elements = 0; if (max_detections > 0 && max_classes_per_detection > 0) { // Check if the product would exceed the maximum size_t if (static_cast(max_detections) > std::numeric_limits::max() / static_cast(max_classes_per_detection)) { throw std::runtime_error("Integer overflow in max_detections * max_classes_per_detection"); } else { num_elements = static_cast(max_detections) * max_classes_per_detection; } } ``` This fix checks for potential integer overflow before performing the multiplication, ensuring safe memory allocation.