I was making a Firefox Extension to quickly check my account balance at one of the banks I use. They have some annoying password policies 6-8 characters [A-Za-z0-9] but they do lock out an account after a couple of failed attempts and require out of band authentication (SMS). So after I finished what should have worked, the response I was getting was constantly bad password or client ID. After some digging I found the password submitted kept changing each time. (I didn't know it wasn't my password the first time, because who in the right mind doesn't use a password manager)

After a few moments of looking over the included scripts I came across this in the JavaScript.

//p is the real password
function NABcrypt(p,k,a) {
//Looks for repeated chars
  for (var i=a.length-1;i>0;i--) {
    if (i!=a.indexOf(a.charAt(i))) {
	  a=a.substring(0,i)+a.substring(i+1);
	}
  }
  var r=new Array(p.length);
  for (var i=0;i<p.length;i++) {
    r[i]=p.charAt(i);
    var pi=a.indexOf(p.charAt(i));
    if (pi>=0 && i<k.length) {
      var ki=a.indexOf(k.charAt(i));
      if (ki>=0) {
	    pi-=ki;
		if (pi<0) pi+=a.length;
	    r[i]=a.charAt(pi);
	  }
    }
  }
  return r.join("");
}

A simple key based subistution cipher k changes and a is 0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ.

I've only come across something similar before in aSSL (a broken by design transport security). I can't for the life of me think of why this would be useful, if it is to protect against client side malware, a simple software keylogger would already have the password.

Please don't do this.

I felt Responsible Disclosure did not apply as there is no risk, transport layer security is implemented.