arXiv blog

Quantum Noise Breaks Random Number Generator Record

The quantum noise in a laser beam has been used to generate random numbers at the rate of 300 Mbits per second, breaking a record that stood for just a few days

kfc 06/21/2010

  • 6 Comments

Random numbers are useful beasts, in particular for cryptographers who use them to generate their codes.

But how best to make random numbers at useful speeds? The question is intimately linked to the nature of randomness. One way is to rely on a computer to do the task using an algorithm that generates numbers that look and feel random. For example, the digits of pi appear random but can be generated by a simple algorithm.

Mathematicians treat these so-called pseudo random numbers with caution. One reason they value random numbers is precisely because they cannot be generated by any logical process: anything that can be calculated, like pi, clearly falls outside this category. Cryptographers, in particular, want codes that cannot be calculated by an eavesdropper with a powerful enough computer.

The current consensus is that the best source of randomness is the quantum world where uncertainty rules. We recently looked at just how good this randomness is. Any uncertainty in the value of physical quantities such as position, momentum, electric field will do.

One way of exploiting this uncertainty is to send photons towards a semi-silvered mirror and measure whether they are reflected or transmitted. The resulting sequence of 0s and 1s ought to be truly random.

The problem here is the rate at which truly random numbers can be generated. Commercial devices that work in this way with semi-silvered mirrors generate random numbers at the rate of up 16 Mbits per second.

Today, Yu Liu and buddies at Peking University in China reveal a way to dramatically speed this up. Their idea is to use the phase noise in a laser to generate random numbers.

Phase noise is the result of quantum uncertainties in the way photons are spontaneously emitted inside a laser--due to vacuum fluctuations of the electric field, for example. The causes the photons to be emitted with frequencies which differ very slightly.

Liu and co use this small change in frequency to create a beat signal by delaying a portion of the laser light and interfering it with the later output. They then sample the beat signal with an analogue to digital converter.

After a little mathematical cleaning, the resulting sequence of numbers passes a number of standard tests of randomness and Liu and friends are able to pronounce it truly random.

What's impressive about this technique is the data rate they achieve. Liu and co are claiming data rates of up to 300 Mbits per second. And since the vertical cavity laser that they use for this can be relatively easily integrated into modern electronic circuitry, there's nothing to stop it being used almost immediately for commercial applications

That's not the fastest random number generator--that record goes to an Israeli team that recently achieved 3 Gbits per second using the chaotic output of a laser. That's good but not quantum-based randomness, which is the gold standard for many physicists.

Earlier this month, a team in the US claimed the record for the fastest quantum generated random number at 100 Mbits per second, based on the intervals between photon arrivals at a detector.

Now, just a few days later, the Chinese have done even better. That's should set in train an interesting numbers battle with other providers of random numbers over who can generate randomness of the highest quality the fastest. Prepare for battle!

Ref: arxiv.org/abs/1006.3512: Truly Random Number Generation Via Entropy Amplification

TRSF: Read the Best New Science Fiction inspired by today’s emerging technologies.

Print

Close Comments

To comment, please sign in or register

Forgot my password

genenphotos

1 Comment

  • 602 Days Ago
  • 06/21/2010

randomness tests

Time was, pi passed all known tests for randomness.  Must be new randomness tests out there.

Reply

momsenpe

1 Comment

  • 602 Days Ago
  • 06/21/2010

Re: randomness tests

This is an interesting article; however, I couldn't get passed the many grammatical mistakes. That aside, I'm enjoying the race for the best randomness generator. GO USA!!

Reply

bja009

1 Comment

  • 578 Days Ago
  • 07/15/2010

Re: randomness tests

Couldn't get 'passed' them, huh?
/facepalm

Reply

Silentvoice

1 Comment

  • 602 Days Ago
  • 06/21/2010

Re: randomness tests

I believe that numerical evidence suggests that the digits of pi are uniformly distributed. I don't think that a strong theorem concerning its actual 'randomness' has been established yet.

Reply

ilovestevi

1 Comment

  • 567 Days Ago
  • 07/26/2010

Random Number Generation Speed

Comments by Leonard Rarick (leonard.rarick@gmail.com) on
http://www.technologyreview.com/blog/arxiv/25355/

The report from Yu Liu of a true random number generation rate of 300 Mbits per second from quantum mechanical processes is an important contribution.  I agree that “quantum based randomness … is the gold standard” for random number generation.  And the 3 Gbit rate from chaotic laser output is also an important true random number method.  However, this is not the fastest generation method.  The Rock processor chip [1] developed at Sun Microsystems, now Oracle, provides a random number generator with a higher rate.  The Rock processor chip random number generator [2, 3] uses a combination of true random number sources (8 ring oscillators of relative prime lengths scattered over the chip), 24 practically strong random number generator sources [4], and hashing these sources with 348 bits of internal state each clock, resulting in 64 output random bits each clock.  The clock rate is 2.3 GHz, resulting in over 147 Gbits per second.  The Rock chip contains 16 cores and is available to groups doing research in parallel processing.  A random number generator is provided for each pair of cores, so there are eight random number generators on the chip.  Eight instantiations of any random number generator will provide eight times the output, nevertheless, the total number of random number bits produced by a Rock chip is in excess of one trillion bits per second.  If desired, Liu’s quantum method or the chaotic laser method could be used as the true random generation source instead of the ring oscillators that were used in the Rock implementation.

[1] Konstadinidis, et.al., “Architecture and Physical Implementation of a Third Generation 65nm,. 16 Core, 32 Thread Chip-Multithreading SPARC Processor”, IEEE Journal of Solid State Circuits, 2009, vol. 44, Issue 1, pages 7-17
[2] Patent number 7,233,065: Continuous Random Number Generation Method and Apparatus.
[3] Patent number 7,082,449: Method and Apparatus for Generating Pseudo-Random Numbers.
[4] Peter Gutmann, “Software Generation of Practically Strong Random Number”, presented at the 7th USENIX Security Symposium, 1998.

Reply

AlekseyGalan

1 Comment

  • 493 Days Ago
  • 10/08/2010

Review articles: the quantum random number generators

I did a review of recent articles in the field of quantum random number generators.
Follow this link: http://scisatellite.blogspot.com/2010/10/review-articles-quantum-random-number.html

Reply

Bio

The Physics arXiv Blog produces daily coverage of the best new ideas from an online forum called the Physics arXiv on which scientists post early versions of their latest ideas. Contact me at KentuckyFC @ arxivblog.com

Follow The Physics arXiv Blog on Twitter

Subscribe to the arXiv blog RSS Feed

Advertisement
Advertisement

Facebook

Advertisement