Monday, December 7, 2015


I'm fooling around with a new Swift Cocoa project. The Encoder app does encryption and decryption. I use the phrase "fooling around" deliberately because an important rule of cryptography is to use standard libraries. If you don't really know what you're doing, you certainly shouldn't "roll your own."

Here is a screenshot of the current version of the app.

At the heart of it, we are doing something trivial:

return Zip2Sequence(a1,a2).map { $0^$1 }

where a1 and a2 are binary data, i.e. arrays of integers with values between 0 and 255 (UInt8). The xor operator ^ is just what we want for both encoding and decoding.

The trick, of course, is to generate a keystream of pseudo-random numbers given a key, which is a String.

Foundation provides standard functions for obtaining random numbers:
rand, random, arc4random, arc4random_uniform.

arc4random and arc4random_uniform are preferred for really random PRNG, but lack the ability to be seeded (by the caller). Seeding is needed so that the same pseudo-random keystream sequence can be regenerated, given a relatively short key.

The rand function returns an Int32 (try rand() is Int32 in a Swift playground. So in theory it might return both positive and negative values. However, it only seems to return positive numbers, which is good, because, say -3 % 256 is equal to -3, which would cause trouble when trying to convert to UInt8. According to this (very old) file, the OS X rand uses unsigned values.

The maximum value returned is RAND_MAX, which is equal to 2147483647 = 2^31 - 1, which is also equal to Int32.max. What we want are integers in the interval [0,255], so I just do:

Int(rand()) % 256

I'm not an expert, of course, but I looked at the distribution of output of rand and it seemed sufficiently random for my purposes.

The next question is that of seeding the PRNG. You might do:


The seed function srand takes a UInt32. The question is, how do we turn our key into a UInt32? What I've done so far is to use the String property hashValue / This gives an Int, which is 64 bits, and of course, may also be negative.

Here is my approach:

Having initialized everything, we just call

func next() -> UInt8 {
  return UInt8( Int(rand()) % 256 )

One issue I'm still working on is that of loading binary data. I know how to turn an [UInt8] into data:

let a: [UInt8] = Array(0..<4)

let data = NSData(bytes: a, length: 4)

and load data from a file with NSData(contentsOfFile:fn), but I haven't been able to figure out how to turn that back into an array of UInt8 in Swift.

I puzzled out a way around the problem, however. String will take NSData in an initializer:

Given that, I get the individual bytes from the string and use a Dictionary

As usual these days, the project is on github here. Thanks to samol for explaining how to do the gist thing.


I found a way to do the data conversion mentioned above. It looks like it is probably the natural way to do this in Swift.

We use NSInputStream initialized with the NSData object. We allocate the necessary space:

var buffer = Array(count: n, repeatedValue: 0)

and then read the data into the buffer using a reference., maxLength: n)

The last call returns a result, which is the number of bytes read. Alternatively, one could read byte by byte until stream.hasBytesAvailable returns false. The only thing that confused me for a while was that I had Array(0..<n) which gave Int values of 64 bits each. Without the map on line 4 the data looks like:

[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 2,

which I interpret as 00000000, 10000000, 20000000 and so on. These are the decimal values for each byte, where the ordering is little-endian, with the low value byte having the first memory address.

Intel x86 processors store a two-byte integer with the least significant byte first, followed by the most significant byte. This is called little-endian byte ordering.

Apple docs. If I write the 64-bit data to a file and examine it with hexdump:

> hexdump x.bin
0000000 00 00 00 00 00 00 00 00 01 00 00 00 00 00 00 00

No comments: