I'd expect that all you have to do is read the bytes into your 32-bit int, account for endianness, and the sign bit will be exactly where it ought to be. The way you're describing this is a bit ...