Currently implementation of the test relies on changing the last byte of the cypher to mess up the padding lines. However after reading though the pkcs5 rfc, if the last bit is already 0 the padding will not be taken into account and the result will have no error but will be a garbage output. Current test doesn't take this into account.

The fix should encode the test text without padding and decode with padding to make sure that the bad padding exception is thrown.