In .NET, we have the SecureString class, which is all very well until you come to try and use it, as to (for example) hash the string, you need the plaintext. I've had a go here at writing a function that will hash a SecureString, given a hash function that takes a byte array and outputs a byte array.
private static byte[] HashSecureString(SecureString ss, Func<byte[], byte[]> hash)
{
// Convert the SecureString to a BSTR
IntPtr bstr = Marshal.SecureStringToBSTR(ss);
// BSTR contains the length of the string in bytes in an
// Int32 stored in the 4 bytes prior to the BSTR pointer
int length = Marshal.ReadInt32(bstr, -4);
// Allocate a byte array to copy the string into
byte[] bytes = new byte[length];
// Copy the BSTR to the byte array
Marshal.Copy(bstr, bytes, 0, length);
// Immediately destroy the BSTR as we don't need it any more
Marshal.ZeroFreeBSTR(bstr);
// Hash the byte array
byte[] hashed = hash(bytes);
// Destroy the plaintext copy in the byte array
for (int i = 0; i < length; i++) { bytes[i] = 0; }
// Return the hash
return hashed;
}
I believe this will correctly hash the string, and will correctly scrub any copies of the plaintext from memory by the time the function returns, assuming the provided hash function is well behaved and doesn't make any copies of the input that it doesn't scrub itself. Have I missed anything here?
See Question&Answers more detail:
os 与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…