BLOG

Security is a matter of numbers: encryption, signature and authentication protocols are all based on numbers and normally the keys used for these protocols are generated on the basis of a seed that must be random. If someone could guess the seed of your crypto schema, it would be really easier to reconstruct the related keys. So what? It's only a random number, choose one randomly and that's all. Unfortunately, despite what one could argue, the identification of a true randomly generated number is not an easy task for a computer.

Today there exist software solutions able to generate what we could call pseudo-random-generated numbers, which statistically  provide a good approximation of a random process, however the generation results costly in term of computational resources, limiting the possibilities of the massive use of one time keys and limiting also the security level of the generated key; and as we are talking about software, the process might be easily corrupted. On the other hand, so far dedicated hardware for generating random numbers was out of the budget (and capacities) of basic computer users: for for this reason, the news released in the last days from Intel was seen by the technical community as a real storm on the future security panorama.

Intel® recently made public a new RdRand instruction (see Intel Advanced Vector Extensions Programming Reference, Chapter 7 (http://software.intel.com/en-us/avx/)), supported by a new Digital Random Number Generator (DRNG), to be available in future Intel processors. This generator is completely digital and can generate true random numbers at a very high rate.
From a technical point of view, with respect to the RNG taxonomy, Intel DRNG follows the Cascade Construction RNG model, using a processor resident entropy source to repeatedly seed a hardware-implemented Cryptographically Secure Pseudo-Random Number Generator (CSPRNG).

With respect to traditional software approaches, it has some notable advantages:

1. It includes an entropy source implementation allowing to seed the CSPRNG with high quality entropy;
2. It is a self-contained hardware module, completely isolated from software attacks on its internal state;
3. It provides statistical quality (independence and uniform distribution), highly unpredictable random number sequences and high performance.

But what makes this news sensational is the fact that this DRNG process is implemented in hardware on the processor silicon itself and can be utilized through a new instruction added to the Intel® 64 instruction set allowing software with the proper permissions to bypass intermediate software stacks, libraries or operating system layer, consequently (at least theoretically) saving time in the generation of chains of DRNG.

If, as it appears, Intel will put on the market a new line of microprocessor equipped with this new instruction set and hardware module, it is easy to foresee the begin of a new era for security applications, which will have the "hands" plenty of fast and low cost random numbers!