Java CipherInputStream在解密时的行为

Question

I am using the following code to decrypt files encrypted from android device.

private void mDecrypt_File(FileInputStream fin, String outFile) throws Exception {
  FileOutputStream fout = new FileOutputStream(outFile);

  byte[] iv = new byte[16];
  byte[] salt = new byte[16];
  byte[] len = new byte[8];
  byte[] FC_TAGBuffer = new byte[8];

  Cipher cipher = Cipher.getInstance(CIPHER_INSTANCE);

  DataInputStream dis = new DataInputStream(fin);

  dis.read(iv, 0, 16);
  dis.read(salt, 0, 16);

  Rfc2898DeriveBytes rfc = new Rfc2898DeriveBytes(DEFAULT_PASSWORD, salt, F_ITERATIONS);
  SecretKey key = new SecretKeySpec(rfc.getBytes(32), "AES");

  //decryption code
  cipher.init(Cipher.DECRYPT_MODE, key, new IvParameterSpec(iv));
  CipherInputStream cIn = new CipherInputStream(dis, cipher);

  cIn.read(len, 0, 8);
  long lSize = getLong(len, 0);

  cIn.read(FC_TAGBuffer, 0, 8);

  byte[] tempFC_TAGBuffer = changeByteArray(FC_TAGBuffer, 0);//new byte[8];                           

  BigInteger ulong = new BigInteger(1, tempFC_TAGBuffer);

  if (!ulong.equals(FC_TAG)) {
    Exception ex = new Exception("Tags are not equal");
    throw ex;
  }

  byte[] bytes = new byte[BUFFER_SIZE];
  //determine number of reads to process on the file                          
  long numReads = lSize / BUFFER_SIZE;
  // determine what is left of the file, after numReads                   
  long slack = (long) lSize % BUFFER_SIZE;

  int read = -1;
  int value = 0;
  int outValue = 0;

  MessageDigest md = MessageDigest.getInstance("SHA-256");
  md.reset();
  // read the buffer_sized chunks         
  for (int i = 0; i < numReads; ++i) {
    read = cIn.read(bytes, 0, bytes.length);
    fout.write(bytes, 0, read);
    md.update(bytes, 0, read);
    value += read;
    outValue += read;
  }
  // now read the slack                   
  if (slack > 0) {
    read = cIn.read(bytes, 0, (int) slack);
    fout.write(bytes, 0, read);
    md.update(bytes, 0, read);
    value += read;
    outValue += read;
  }
  fout.flush();
  fout.close();
  byte[] curHash = md.digest();

  byte[] oldHash = new byte[md.getDigestLength()];
  read = cIn.read(oldHash, 0, oldHash.length);
  if (oldHash.length != read || (!CheckByteArrays(oldHash, curHash))) {
    Exception ex = new Exception("File Corrupted!");
    throw ex;
  }
  if (outValue != lSize) {
    Exception ex = new Exception("File Sizes don't match!");
    throw ex;
  }
}

This code is working fine on android but behaving strange on Java desktop application. What I have observed is, on reading old hash from CipherInputStream cIn returns correct hash value only if the size of data to be decrypted is multiples of 32. For example, if I encrypt a text file which have a text of length 32 chars(or 64/128/...), then the following code

byte[] oldHash = new byte[md.getDigestLength()];
read = cIn.read(oldHash, 0, oldHash.length);
if (oldHash.length != read || (!CheckByteArrays(oldHash, curHash))) {
    Exception ex = new Exception("File Corrupted!");
    throw ex;
}

calculates oldHash correctly, but if I change the text of any other length(not multiple of 32) then the oldHash's last few values becomes zeros.

My Observations :

1. Text Size 6 char - Trailing zeros in oldHash - 6
2. Text Size 13 char - Trailing zeros in oldHash - 13
3. Text Size 20 char - Trailing zeros in oldHash - 4
4. Text Size 32 char - Trailing zeros in oldHash - 0 // Correct Result
5. Text Size 31 char - Trailing zeros in oldHash - 1
6. Text Size 64 char - Trailing zeros in oldHash - 0 // Correct Result

Please help me understanding this behavior.

Answer 1

Agree with DuncanJones, your loop is a mess. Although you properly check the return value of the read() method your loop iterations assume that every read() will return BUFFER_SIZE bytes or 'slack' bytes for the last read.

You code would be hugely better if you make proper use of DataInputStream . For example, you wrap the FileInputStream fin in a DataInputStream but then use the wrong methods in these two lines:

dis.read(iv, 0, 16);
dis.read(salt, 0, 16);

instead, you should use the readFully method, as in:

dis.readFully(iv);
dis.readFully(salt);

Similarly, you would benefit from wrapping your CipherInputStream cIn with another DataInputStream, something like:

CipherInputStream cIn = new CipherInputStream(dis, cipher);
DataInputStream dcIn = new DataInputStream(cIn);

DataInputStream already has a getLong method, so you could just replace these lines:

cIn.read(len, 0, 8);
long lSize = getLong(len, 0);

cIn.read(FC_TAGBuffer, 0, 8);

with

long lSize = dcIn.getLong()
dcIn.readFully(FC_TAGBuffer);

and you get to throw out your homegrown getLong method. Now you can go on and read the next lSize bytes in exactly BUFFER_SIZE chunks using dcIn.readFully(bytes) and make your code, cleaner, shorter, easier to read, and correct.