The vulnerability described is an inefficient decoding of ASCIIHexDecode streams in pypdf. The provided information included a link to a pull request (PR #3666) that patches the vulnerability. By inspecting the commits within this pull request, specifically commit a7f4df49c0af8aac582101a5c466d82688a029ee, I was able to pinpoint the exact code change that addresses the issue. The commit modifies the decode function within the ASCIIHexDecode class in the pypdf/filters.py file. The original, vulnerable code iterated through the data byte-by-byte, which is inefficient. The patch replaces this with a much more performant implementation that uses the built-in binascii.unhexlify function. Therefore, the ASCIIHexDecode.decode function is the vulnerable function.