From Dave Farber’s Interesting People mailing list, Ethan Ackerman writes: “The SANS blog has a nice summary of an innovative exploit in MS
Vista’s speech recognition implementation – a malicious audio file can
be used to give user-level control. It does so not in the usual
buffer overflow or misleading header corruptions, but by actual voice
recognition of spoken scripting commands.
Playing a sound file featuring spoken commands to open a malicious website.”