Cloudflare translates photos of 100 lava lamps into random data for use in SSL encryption.
After reading this article you will be able to:
Copy article link
Randomness is extremely important for secure encryption. Each new key that a computer uses to encrypt data must be truly random, so that an attacker won't be able to figure out the key and decrypt the data. However, computers are designed to provide predictable, logical outputs based on a given input. They aren't designed to produce the random data needed for creating unpredictable encryption keys.
To produce the unpredictable, chaotic data necessary for strong encryption, a computer must have a source of random data. The "real world" turns out to be a great source for randomness, because events in the physical world are unpredictable.
As one might expect, lava lamps are consistently random. The "lava" in a lava lamp never takes the same shape twice, and as a result, observing a group of lava lamps is a great source for random data.
To collect this data, Cloudflare has arranged about 100 lava lamps on one of the walls in the lobby of the Cloudflare headquarters and mounted a camera pointing at the lamps. The camera takes photos of the lamps at regular intervals and sends the images to Cloudflare servers. All digital images are really stored by computers as a series of numbers, with each pixel having its own numerical value, and so each image becomes a string of totally random numbers that the Cloudflare servers can then use as a starting point for creating secure encryption keys.
Thus, with the help of lava lamps, Cloudflare is able to offer extremely strong (and sufficiently random) SSL/TLS encryption to its customers. This is especially important considering that millions of Internet properties use Cloudflare.
In cryptography, random does not just mean statistically random; it also means unpredictable. Suppose someone were to roll a single six-sided die two dozen times with the following results:
1, 2, 3, 4, 5, 6, 1, 2, 3, 4, 5, 6, 1, 2, 3, 4, 5, 6, 1, 2, 3, 4, 5, 6
Statistically speaking, this is a random distribution of die-rolling results. Each number has an equal probability of being rolled, so it's within the realm of possibility that this sequence would appear.
However, this sequence is not unpredictable. If this series were used in encryption, an attacker could figure out the pattern.
Encrypted data should look like totally random data, since predictable data can be guessed. If there is any pattern – if certain values are used for encryption more than others, or if values appear in a certain order consistently – then mathematical analysis will pick up on the pattern and allow an attacker to have a much easier time guessing the key used for encryption. Essentially, if encrypted data is predictable, it might as well already be compromised.
The process of encryption itself is a predictable one: Encrypted data plus the right key equals decrypted data, and the decrypted data is the same as it was before it was encrypted. But the encryption keys used have to be unpredictable.
To understand why unpredictability is so important, imagine two poker players: Bob always bets when he has good cards and folds (declines to match other players' bets) when he has bad cards. Alice, meanwhile, mixes up her betting strategy so that there's no discernable pattern to it: sometimes she bets when she has good cards, sometimes she contents herself with matching other players' bets, and sometimes she even bluffs by betting big when she has bad cards. When Alice and Bob enter the same poker tournament, Alice lasts much longer than Bob does, because Bob is too predictable. Opponents quickly figure out when Bob has good cards and react accordingly. Even though they can't see his cards, they can discern roughly what cards he's holding.
Similarly, even though attackers can't see the "cards" – or, the encrypted content – that's sent over a network, they can guess it if the method for concealing the content is too predictable.
Computers run on logic. A computer program is based on if-then statements: If certain conditions are met, then perform this specified action. The same input into a program results in the same output every time.
This is by design. An input should lead to an expected output, not an unexpected one. Imagine the chaos if a printer printed random text that was different from the text in the document that was sent to the printer, or if smartphones were to call a different phone number than the one the user entered. Computers are only useful because of their (relative) reliability and predictability. However, that predictability is a liability when it comes to generating secure encryption keys.
Some computer programs are good at simulating randomness, but not good enough at it for creating encryption keys.
A software program called a pseudorandom number generator (PRNG) is able to take an unpredictable input and use it to generate unpredictable outputs. Theoretically, a PRNG can produce unlimited random outputs from a random input.
Such an algorithm is called "pseudorandom" and not "random" because its outputs are not actually completely random. Why is this the case? There are 2 main reasons:
Because of reason No. 2, the algorithm continually needs new inputs of randomness. A random input is known as a "cryptographic seed."
A cryptographically secure pseudorandom number generator, or CSPRNG, is a PRNG that meets more stringent standards, making it safer to use for cryptography. A CSPRNG meets two requirements that PRNGs may not necessarily meet:
Like a PRNG, a CSPRNG needs random data (the cryptographic seed) as a starting point from which to produce more random data.
To generate encryption keys for SSL/TLS encryption, Cloudflare uses a CSPRNG, with data collected from the lava lamps as part of the cryptographic seed.
A cryptographic seed is the data that a CSPRNG starts with for generating random data. Although a CSPRNG could theoretically produce unlimited random outputs from a single cryptographic seed, it is far more secure to regularly refresh the cryptographic seed. An attacker may eventually compromise the initial cryptographic seed – and remember, a CSPRNG will produce the exact same outputs again if it is fed the same seed, so the attacker could then duplicate the random outputs. Additionally, even the most rigorously tested CSPRNG is not guaranteed to produce unpredictable results indefinitely.
With the lava lamps, Cloudflare has a continual source for new cryptographic seed data. Each image the camera takes of the lamps is different, resulting in a different random sequence of numerical values that can be used as a seed.
Many operating systems have their own sources of random data for use in cryptographic seeds, for instance from user actions (mouse movements, typing on a keyboard, etc.), although they obtain this data relatively slowly. Cloudflare mixes the random data obtained from the lava lamps with data generated by the Linux operating system on two different machines in order to maximize entropy when creating cryptographic seeds for SSL/TLS encryption.
In general, "entropy" means disorder or chaos. But entropy has a specific meaning in cryptography: it refers to unpredictability. Cryptographers will actually measure how much entropy a given set of data has in terms of the number of bits of entropy. Because of this, Cloudflare refers to the lava lamp wall as the "Wall of Entropy."
Because the lava lamp wall is in the busy lobby of the Cloudflare headquarters, this happens all the time. People come and go in the lobby, walking by or stopping to talk in front of the lamps. Such obstructions become part of the randomness that the camera captures, so people partially blocking the camera's view of the lava lamps actually helps generate entropy.
If this happens, Cloudflare still has two other sources for randomization from the Linux operating system running on Cloudflare servers. In addition, Cloudflare has easy physical access to the camera because it's in a Cloudflare-owned space, and Cloudflare can quickly turn it back on or replace it as needed.
The other two main Cloudflare offices are in London and Singapore, and each office has its own method for generating random data from real-world inputs. London takes photos of a double-pendulum system mounted in the office (a pendulum connected to a pendulum, the movements of which are mathematically unpredictable). The Singapore office measures the radioactive decay of a pellet of uranium (a small enough amount to be harmless).
Surprisingly, no – a company called Silicon Graphics designed a similar system called "Lavarand" in 1996, although the patent has since expired.
To learn more about the Cloudflare lava lamp wall check out these two blog posts:
Learning Center Navigation