An integer overflow in the bounds verification of a 2D matrix kernel in the imageproc library allows crafted kernel objects to cause out-of-bounds reads. This occurs because the total size check casts after multiplication, failing to detect overflow. The flaw affects multiple versions prior to fixed releases and can lead to unsafe memory access.
The imageproc library contains an out-of-bounds read vulnerability caused by improper handling of NaN coordinates during bilinear and bicubic sampling. This flaw allows an attacker to craft floating point coordinates that bypass bounds checks, leading to arbitrary reads of memory, including the first 32 bits of the address space with bilinear sampling and a few bytes beyond allocated memory with bicubic sampling. Multiple versions prior to fixed releases are affected.
The imageproc library contains a vulnerability in its pixel sampling bounds check. The check attempts to constrain coordinates within image bounds by adding a constant and taking the minimum with dimension-1, but this approach does not prevent integer overflow. As a result, malicious inputs can cause out-of-bounds sampling at multiple coordinates.
The imageproc library contains a vulnerability in its bounds checking when sampling from images. The flaw arises because a floating-point bounds check improperly handles NaN values, allowing out-of-bounds memory reads. This can lead to arbitrary reads in low memory addresses when using bilinear sampling, and potentially reads beyond allocated memory with bicubic sampling. The issue affects multiple versions prior to fixed releases.
A vulnerability in the imageproc library involves improper bounds checking due to an overflow in the calculation of slice storage size for a 2D matrix kernel. This flaw allows crafted kernel objects to cause out-of-bounds access in unsafe code, potentially leading to memory safety violations when library functions trust these preconditions.
The imageproc library contained a vulnerability due to a fragile bounds check when sampling pixels from an image. The code attempted to constrain coordinates within image bounds by adding a constant to an input coordinate and taking the minimum with 'dimension - 1', but this approach did not prevent integer overflow. As a result, malicious inputs could cause coordinate calculations to exceed image bounds, allowing out-of-bounds pixel reads.