AIKIDO-2026-817514

TL;DR

pypdf extracts inline images from page content streams using dedicated decoders for the ASCII85 and ASCIIHex filters. These decoders read the stream in a loop and rewind two bytes whenever a terminator is not yet found, but they fail to detect when the underlying stream is already exhausted. A crafted PDF whose inline image is never properly terminated makes the decoder rewind and re-read the same bytes endlessly, hanging operations such as page text extraction. The fix detects an empty read from the stream and raises a read error instead of looping forever.