Question

This is not homework. I would like to generate a random integer sequence (50 digits between 0:9) from /dev/random in R. I have a hardware entropy key.

I have found two “ideas”, neither of which I can get to supply me with the numbers I am after:

1) RDieHarder. Seems to allow access to /dev/random, but I cannot get it to produce the integer sequence that I require. e.g.

>library(RDieHarder)  
>x  <-dieharder(rng="/dev/urandom", psample=50) #urandom used for example

2) The accuracy package can supply true random numbers, but appears to be out of date and I can’t see how to just sequence from /dev/random. e.g.

>library(accuracy)
>x=runifT(50)

Yes I have read Knuth etc and understand the problems of TRNGs (hence the hardware entropy key).

Any other ideas? Thanks.

Answer 1

Here’s how to read from a dev and get n numbers from a to b inclusive:

readRandom <- function(n,a,b,dev="/dev/urandom"){
  size = b-a + 1
  rng = file(dev,"rb") # open connection
  nums = readBin(rng,what="integer",n=n) # read some 8-byte integers 
  close(rng) # close the connection
  return( a + nums %% size ) # reduce range and shift
}

If I read from /dev/random it blocks since my system runs out of entropy, but your key should be busy feeding entropy into the system, I think….

The ‘edge’ effect problem is as follows. Suppose I generate random integers from 0 to 10, but you want integers from 0 to 6. Then X %% 7 would generate twice as many 0,1,2,3 values, because the mapping is this:

> (0:10) %% 7
 [1] 0 1 2 3 4 5 6 0 1 2 3

Now readBin is getting 8-bit integers, which are huge, so the odd few extra numbers at the end of the sequence shouldn’t make much difference…