First, consider that you may be reducing a huge string to a small fixed-length digest. That means that information got lost; there may be many long strings that also map to the exact same digest. However, if the digest works properly, it should be very computationally expensive to figure out a plain text value that results in a given digest.