LETTERS

Securing Secure Algorithms

Bruce Schneier's article, "The Cambridge Algorithms Workshop" (DDJ, April 1994) lists three conditions under which an encryption algorithm should be secure. There is one class of encryption algorithms that not only satisfies these conditions, but has been mathematically proven to be impossible to break--those algorithms that allow the use of a key that is longer than the plaintext to be encrypted. In addition to being absolutely secure, these algorithms are the simplest to implement and the fastest to execute. The proof of unbreakability is quite simple and based on the following theorem.

Theorem: Any binary number N, when XORed with a random binary number R, produces a random number.

Proof:

Let K be a random sequence of X number of bytes.

Let P be plaintext of length X.

Let C be the ciphertext produced by K xor P.

From the theorem we know that the ciphertext C is a sequence of totally random bytes and is therefore unbreakable.

Example 1 shows P1 and P2 as plaintext to be encrypted. K1 is a random series of 8-bit numbers used to encrypt P1. C1 is the ciphertext produced by XORing P1 with K1. C2 is the ciphertext produced by using C1 as the key when encrypting P2.

If ciphertext C1 is unencrypted by XORing it with the random sequence K1, then we get plaintext P1. If ciphertext C1 is unencrypted by XORing it with cipher- text C2, then we get plaintext P2! Thus we see that, given any plaintext P of length ten, there exists a key that will produce P when the key is XORed with C1. That is, C1 could be unencrypted into any ten-character message, given the right random key.

One criticism of this type of data-encryption algorithm is that to encrypt ten megabytes of plaintext you need a random-number key that is ten megabytes long. In this age of CDs and laser disks, however, distributing large keys to those with whom you correspond is no more difficult than distributing short keys.

William D. Hause

Bryan, Texas

The simplicity of this method, shown in Example 2, highlights deficiencies in the SDK documentation (which is improving slowly). The only use of this method I could find was in the C7 CDDEMO sample program.

On the topic of minimizing modeless dialog boxes, I find that Joseph's method seems to have some redundant processing--notably the setting of the dialog icon word to NULL. In my experiments, the dialog-icon word is always NULL. The only reason I can see that he might need this is if some other application running has changed the dialog-icon word. This is a bit naughty--since all applications share the dialog class, no application should be changing the dialog-class word. Joe's approach then removes the need to process the WM_NCPAINT message and the other places where he sets the dialog-class word to NULL.

Minimizing the dialog works fine if the icon is outside the application window, but not if the icon is over the application's client area. This is because the icon's background is always the desktop bitmap, rather than the windows underlying the icon. We need to get windows beneath the icon to repaint before using DrawIcon, rather than the simple use of WM_ICONERASEBKGND. I can't help feeling that Windows should have already repainted any underlying windows, and the application should not be forcing the background to repaint anyway! Since modeless dialogs with icons are a requirement, maybe we need to ask Microsoft to provide these in a future version of Windows.

David Lowndes

Stoke-on-Trent, England

Far Be It From Us

Dear DDJ,

Multimedia World and PC Magazine have decided to no longer accept advertising for erotic, pornographic "adult" multimedia or CD-ROM products.

Degrading and dehumanizing women into sex objects both contribute to physical violence against women and have been invaluable assets in the war against social, political, and economic equality for women.

The constitution protects the right of free speech. It does not guarantee the right to make a profit. I am writing to ask you to take a positive step toward promoting social, political, and economic equality for women by refusing to accept "adult software" advertising in your magazine. For equality and safety.

Lisa Gray

Bothell, Washington

DDJ responds: Thanks for writing; we appreciate hearing from you. While we agree in principle with your statements, we're curious about what led you to believe that Dr. Dobb's Journal accepts advertising for erotic products?

By the way, the constitution does, in fact "guarantee the right to make a profit," unless we missed some important news recently.

Trying Tries

Dear DDJ,

I was happy to see Tom Swan's column on tries (DDJ, April 1994). Though I've been using tries to do many things (for at least ten years) I find that they are misunderstood in general by people who otherwise use binary trees and the like. And so they don't get much press (or use).

I find that for dealing with words, concordance programs and things like that, I almost always put one character in a node with downward links for the next character in the word, and also with sideways links for a similar word but with a different character in this location.

The diagram in Figure 1 shows a trie, which includes the words and, arts, and ant. Each leaf is a null, signifying end of word. This makes for a compact and easily searchable dictionary. I also add an additional item on each leaf, a number, which can be used as a counter. Thus, when a word is inserted into the trie, a count of the number of occurrences can be easily kept.

I also want to thank Tom for using pseudocode in his column. Far too often nowadays people will write a program and call it an "algorithm.'' They don't understand the difference between an algorithm and an implementation, which is something that I found disturbing, since I have been a Modula-2 user for many years and that distinction is so important to the language (and to Professor Wirth himself).

I've read of a time when Algol was the pseudocode of choice. I remember a time when pseudocode really was a readable nonimplementation (those were the days), and now it seems that C has become the pseudocode (which is a sad statement on the state of computer science).

John Andrea

Antigonish, Nova Scotia

Figure 1: Trying tries. Example 1: Securing secure algorithms.

Example 2: Modeless dialog boxes.

/* Global variable - the active dialog handle */
HWND ghActiveDlg = NULL;
case WM_ACTIVATE:
  if ( wParam == WA_INACTIVE )
  {
    /* Becoming inactive */
    ghActiveDlg = NULL;
  }
  else
  {
    /* Becoming active */
    ghActiveDlg = hDlg;
  }
  return( FALSE );
In the main message loop the modeless dialog processing is simply:
/* If one of the modeless dialogs is active process the message */
if ( ( ghActiveDlg == NULL ) || !IsDialogMessage( ghActiveDlg, &msg ) )
{
  /* Otherwise process it as normal */
  TranslateMessage( &msg );
  DispatchMessage( &msg );
}

LETTERS

Securing Secure Algorithms

Rx for Phone Woes

Software Testing

Modeless Dialog Boxes

Far Be It From Us

Trying Tries

Example 2: Modeless dialog boxes.