I am generating a 2048-bit safe prime for a Diffie-Hellman-type key, p such that p and (p-1)/2 are both prime.
How few iterations of Rabin-Miller can I use on both p and (p-1)/2 and still be confident of a cryptographically strong key? In the research I've done I've heard everything from 6 to 64 iterations for 1024-bit ordinary primes, so I'm a little confused at this point. And once that's established, does the number change if you are generating a safe prime rather than an ordinary one?
Computation time is at a premium, so this is a practical question - I'm basically wondering how to find out the lowest possible number of tests I can get away with while at the same time maintaining pretty much guaranteed security.
Let's assume that you select a prime p by selecting random values until you hit one for which Miller-Rabin says: that one looks like a prime. You use n rounds at most for the Miller-Rabin test. (For a so-called "safe prime", things are are not changed, except that you run two nested tests.)
The probability that a random 1024-bit integer is prime is about 1/900. Now, you do not want to do anything stupid so you generate only odd values (an even 1024-bit integer is guaranteed non-prime), and, more generally, you run the Miller-Rabin test only if the value is not "obviously" non-prime, i.e. can be divided by a small prime. So you end up with trying about 300 values with Miller-Rabin before hitting a prime (on average). When the value is non-prime, Miller-Rabin will detect it with probability 3/4 at each round, so the number of Miller-Rabin rounds you will run on average for a single non-prime value is 1+(1/4)+(1/16)+... = 4/3. For the 300 values, this means about 400 rounds of Miller-Rabin, regardless of what you choose for n.
So if you select n to be, e.g., 40, then the cost implied by n is less than 10% of the total computational cost. The random prime selection process is dominated by the test on non-primes, which are not impacted by the value of n you choose. I talked here about 1024-bit integers; for bigger numbers the choice of n is even less important since primes become sparser as size increases (for 2048-bit integers, the "10%" above become "5%").
Hence you can choose n=40 and be happy with it (or at least know that reducing n will not buy you much anyway). On the other hand, using a n greater than 40 is meaningless, because this would get you to probabilities lower than the risk of a simple miscomputation. Computers are hardware, they can have random failures. For instance, a primality test function could return "true" for a non-prime value because a cosmic ray (a high-energy particle hurtling through the Universe at high speed) happens to hit just the right transistor at the right time, flipping the return value from 0 ("false") to 1 ("true"). This is very unlikely -- but no less likely than probability 2-80. See this stackoverflow answer for a few more details. The bottom line is that regardless of how you make sure that an integer is prime, you still have an unavoidable probabilistic element, and 40 rounds of Miller-Rabin already give you the best that you can hope for.
To sum up, use 40 rounds.
The paper Average case error estimates for the strong probable prime test by Damgard-Landrock-Pomerance points out that, if you randomly select k-bit odd number n and apply t independent Rabin-Miller tests in succession, the probability that n is a composite has much stronger bounds.
In fact for 3 <= t <= k/9 and k >= 21,
For a k=1024 bit prime, t=6 iterations give you an error rate less than 10^(-40).
Each iteration of Rabin-Miller reduces the odds that the number is composite by a factor of 1/4.
So after 64 iterations, there is only 1 chance in 2^128 that the number is composite.
Assuming you are using these for a public key algorithm (e.g. RSA), and assuming you are combining that with a symmetric algorithm using (say) 128-bit keys, an adversary can guess your key with that probability.
The bottom line is to choose the number of iterations to put that probability within the ballpark of the other sizes you are choosing for your algorithm.
[update, to elaborate]
The answer depends entirely on what algorithms you are going to use the numbers for, and what the best known attacks are against those algorithms.
For example, according to Wikipedia:
As of 2003 RSA Security claims that 1024-bit RSA keys are equivalent in strength to 80-bit symmetric keys, 2048-bit RSA keys to 112-bit symmetric keys and 3072-bit RSA keys to 128-bit symmetric keys.
So, if you are planning to use these primes to generate (say) a 1024-bit RSA key, then there is no reason to run more than 40 iterations or so of Rabin-Miller. Why? Because by the time you hit a failure, an attacker could crack one of your keys anyway.
Of course, there is no reason not to perform more iterations, time permitting. There just isn't much point to doing so.
On the other hand, if you are generating 2048-bit RSA keys, then 56 (or so) iterations of Rabin-Miller is more appropriate.
Cryptography is typically built as a composition of primitives, like prime generation, RSA, SHA-2, and AES. If you want to make one of those primitives 2^900 times stronger than the others, you can, but it is a little like putting a 10-foot-steel vault door on a log cabin.
There is no fixed answer to your question. It depends on the strength of the other pieces going into your cryptographic system.
All that said, 2^-128 is a ludicrously tiny probability, so I would probably just use 64 iterations :-).
From the libgcrypt source:
/* We use 64 Rabin-Miller rounds which is better and thus
sufficient. We do not have a Lucas test implementaion thus we
can't do it in the X9.31 preferred way of running a few
Rabin-Miller followed by one Lucas test. */
cipher/primegen.c line# 1295
I would run two or three iterations of Miller-Rabin (i.e., strong Fermat probable prime) tests, making sure that one of the bases is 2.
Then I would run a strong Lucas probable prime test, choosing D, P, and Q with the method described here:
https://en.wikipedia.org/wiki/Baillie%E2%80%93PSW_primality_test
There are no known composites that pass this combination of Fermat and Lucas tests.
This is much faster than doing 40 Rabin-Miller iterations. In addition, as was pointed out by Pomerance, Selfridge, and Wagstaff in https://math.dartmouth.edu/~carlp/PDF/paper25.pdf, there are diminishing returns with multiple Fermat tests: if N is a pseudoprime to one base, then it is more likely than the average number to be a pseudoprime to other bases. That's why, for example, we see so many psp's base 2 are also psp's base 3.
A smaller probability is usually better, but I would take the actual probability value with a grain of salt. Albrecht et al Prime and Prejudice: Primality Testing Under Adversarial Conditions break a number of prime-testing routines in cryptographic libraries. In one example, the published probability is 1/2^80, but the number they construct is declared prime 1 time out of 16.
In several other examples, their number passes 100% of the time.
Only 2 iterations, assuming 2^-80 as a negligibly probability.
From (Alfred. J. Menezes. et al. 1996) §4.4 p.148:
Does it matter? Why not run for 1000 iterations? When searching for primes, you stop applying the Rabin-Miller test anyway the first time it fails, so for the time it takes to find a prime it doesn't really matter what the upper bound on the number of iterations is. You could even run a deterministic primality checking algorithm after those 1000 iterations to be completely sure.
That said, the probability that a number is prime after n iterations is 4^-n.
Related
According to the books that i have read, it says that S.H.A(Secure Hash Algorithm) is collision resistant.But if the input space is a 1024 bit number and the output space is a 512 bit message digest then shouldn't it be colliding for
(2^1024)/(2^512) times? As the range is lesser than the domain being mapped there should have been collisions. please explain where i am going wrong.
The chance for a collision does not depend on the input size. The chance to a 512-bit hash collision is 1.4×10^77, see Probability table
Maybe your book has also mentioned the definition of collision resistance? It does not mean that no collisions are created (which is clearly not the case), but that given a hash you are not able to create a message easily that produces this hash.
a hash function H is collision resistant if it is hard to find two
inputs that hash to the same output; that is, two inputs a and b such
that H(a) = H(b), and a ≠ b
From Wikipedia
As you describe: Since the input space (arbitrary size) is larger than the output space (e.g. 512bit for sha512), there always exist collisions.
"Collision resistant" means, it is adequately unlikely for a collision to be found.
Your confusion is answered when considering how large the output space "512 bits" really is:
2^512 (the number of possible configurations of a 512 bit array) is of the order 10^154.
For comparison: The number of atoms in the visible universe is somewhere in the range of 10^80.
A million is 10^6.
So a million of our 'visible universes' has 10^86 atoms.
A million times a million universes has 10^92 atoms.
If you could store a single 512 bit value on a single atom, how many universes would you need to have all possible 512 bit has values stored?
Starting with a specific 512bit number (and assuming the has function is not broken), the probability p to obtain a collision is assuming you can produce new hashes with a rate R and have the total time of t to do this is:
p = R*t/(2^(512/2))
(The exponent is halved, see "birthday attach". The expected search space for a success is to find a collision in n bits is n/2.)
Let's plugin in some example numbers:
The has rate of the bitcoin network is currently about R = 200*10^15 / s (200 million terrahashes per second).
Consider the situation that since the beginning of the universe the bitcoin network's current hashing capacity would have been available for the sole purpose of finding a collision for a specific hash value, i.e. for an available time of t=13.787*10^9 years,
then the probability that a collision would have been found by now is about 7 × 10^-41 %
Again, it is hard to appreciate how small this number is.
Edit: A similar question with a good answer is found here: https://crypto.stackexchange.com/questions/89558/are-sha-256-and-sha-512-collision-resistant
Am I mistaken in thinking that the security of RSA encryption, in general, is limited by the amount of known prime numbers?
To crack (or create) a private key, one has to combine the right pair of prime numbers.
Is it impossible to publish a list of all the prime numbers in the range used by RSA? Or is that list sufficiently large to make this brute force attack unlikely? Wouldn't there be "commonly used" prime numbers?
RSA doesn't pick from a list of known primes: it generates a new very large number, then applies an algorithm to find a nearby number that is almost certainly prime. See this useful description of large prime generation):
The standard way to generate big prime numbers is to take a preselected random number of the desired length, apply a Fermat test (best with the base 2 as it can be optimized for speed) and then to apply a certain number of Miller-Rabin tests (depending on the length and the allowed error rate like 2−100) to get a number which is very probably a prime number.
(You might ask why, in that case, we're not using this approach when we try and find larger and larger primes. The answer is that the largest known prime has over 17 million digits- far beyond even the very large numbers typically used in cryptography).
As for whether collisions are possible- modern key sizes (depending on your desired security) range from 1024 to 4096, which means the prime numbers range from 512 to 2048 bits. That means that your prime numbers are on the order of 2^512: over 150 digits long.
We can very roughly estimate the density of primes using 1 / ln(n) (see here). That means that among these 10^150 numbers, there are approximately 10^150/ln(10^150) primes, which works out to 2.8x10^147 primes to choose from- certainly more than you could fit into any list!!
So yes- the number of primes in that range is staggeringly enormous, and collisions are effectively impossible. (Even if you generated a trillion possible prime numbers, forming a septillion combinations, the chance of any two of them being the same prime number would be 10^-123).
As new research comes out the answer to your question becomes more interesting. In a recent paper "Imperfect Forward Secrecy:How Diffie-Hellman Fails in Practice" by David Adrian et all found # https://weakdh.org/imperfect-forward-secrecy-ccs15.pdf accessed on 10/16/2015 the researchers show that although there probably are a sufficient number of prime numbers available to RSA's 1024 bit key set there are groups of keys inside the whole set that are more likely to be used because of implementation.
We estimate that even in the 1024-bit case, the computations are
plausible given nation-state resources. A small number of fixed or
standardized groups are used by millions of servers; performing
precomputation for a single 1024-bit group would allow passive
eavesdropping on 18% of popular HTTPS sites, and a second group would
allow decryption of traffic to 66% of IPsec VPNs and 26% of SSH
servers. A close reading of published NSA leaks shows that the
agency’s attacks on VPNs are consistent with having achieved such a
break. We conclude that moving to stronger key exchange methods should
be a priority for the Internet community.
The research also shows a flaw in TLS that could allow a man-in-middle attacker to downgrade the encryption to 512 bit.
So in answer to your question there are probably a sufficient quantity of prime numbers in RSA encryption on paper but in practice there is a security issue if your hiding from a nation state.
I have read that it has something to do with time, also you get from including time.h, so I assumed that much, but how does it work exactly? Also, does it have any tendencies towards odd or even numbers or something like that? And finally is there something with better distribution in the C standard library or the Foundation framework?
Briefly:
You use time.h to get a seed, which is an initial random number. C then does a bunch of operations on this number to get the next random number, then operations on that one to get the next, then... you get the picture.
rand() is able to touch on every possible integer. It will not prefer even or odd numbers regardless of the input seed, happily. Still, it has limits - it repeats itself relatively quickly, and in almost every implementation only gives numbers up to 32767.
C does not have another built-in random number generator. If you need a real tough one, there are many packages available online, but the Mersenne Twister algorithm is probably the most popular pick.
Now, if you are interested on the reasons why the above is true, here are the gory details on how rand() works:
rand() is what's called a "linear congruential generator." This means that it employs an equation of the form:
xn+1 = (*a****xn + ***b*) mod m
where xn is the nth random number, and a and b are some predetermined integers. The arithmetic is performed modulo m, with m usually 232 depending on the machine, so that only the lowest 32 bits are kept in the calculation of xn+1.
In English, then, the idea is this: To get the next random number, multiply the last random number by something, add a number to it, and then take the last few digits.
A few limitations are quickly apparent:
First, you need a starting random number. This is the "seed" of your random number generator, and this is where you've heard of time.h being used. Since we want a really random number, it is common practice to ask the system what time it is (in integer form) and use this as the first "random number." Also, this explains why using the same seed twice will always give exactly the same sequence of random numbers. This sounds bad, but is actually useful, since debugging is a lot easier when you control the inputs to your program
Second, a and b have to be chosen very, very carefully or you'll get some disastrous results. Fortunately, the equation for a linear congruential generator is simple enough that the math has been worked out in some detail. It turns out that choosing an a which satisfies *a***mod8 = 5 together with ***b* = 1 will insure that all m integers are equally likely, independent of choice of seed. You also want a value of a that is really big, so that every time you multiply it by xn you trigger a the modulo and chop off a lot of digits, or else many numbers in a row will just be multiples of each other. As a result, two common values of a (for example) are 1566083941 and 1812433253 according to Knuth. The GNU C library happens to use a=1103515245 and b=12345. A list of values for lots of implementations is available at the wikipedia page for LCGs.
Third, the linear congruential generator will actually repeat itself because of that modulo. This gets to be some pretty heady math, but the result of it all is happily very simple: The sequence will repeat itself after m numbers of have been generated. In most cases, this means that your random number generator will repeat every 232 cycles. That sounds like a lot, but it really isn't for many applications. If you are doing serious numerical work with Monte Carlo simulations, this number is hopelessly inadequate.
A fourth much less obvious problem is that the numbers are actually not really random. They have a funny sort of correlation. If you take three consecutive integers, (x, y, z), from an LCG with some value of a and m, those three points will always fall on the lattice of points generated by all linear combinations of the three points (1, a, a2), (0, m, 0), (0, 0, m). This is known as Marsaglia's Theorem, and if you don't understand it, that's okay. All it means is this: Triplets of random numbers from an LCG will show correlations at some deep, deep level. Usually it's too deep for you or I to notice, but its there. It's possible to even reconstruct the first number in a "random" sequence of three numbers if you are given the second and third! This is not good for cryptography at all.
The good part is that LCGs like rand() are very, very low footprint. It typically requires only 32 bits to retain state, which is really nice. It's also very fast, requiring very few operations. These make it good for noncritical embedded systems, video games, casual applications, stuff like that.
PRNGs are a fascinating topic. Wikipedia is always a good place to go if you are hungry to learn more on the history or the various implementations that are around today.
rand returns numbers generated by a pseudo-random number generator (PRNG). The sequence of numbers it returns is deterministic, based on the value with which the PRNG was initialized (by calling srand).
The numbers should be distributed such that they appear somewhat random, so, for example, odd and even numbers should be returned at roughly the same frequency. The actual implementation of the random number generator is left unspecified, so the actual behavior is specific to the implementation.
The important thing to remember is that rand does not return random numbers; it returns pseudo-random numbers, and the values it returns are determined by the seed value and the number of times rand has been called. This behavior is fine for many use cases, but is not appropriate for others (for example, rand would not be appropriate for use in many cryptographic applications).
How does rand() work?
http://en.wikipedia.org/wiki/Pseudorandom_number_generator
I have read that it has something to
do with time, also you get from
including time.h
rand() has nothing at all to do with the time. However, it's very common to use time() to obtain the "seed" for the PRNG so that you get different "random" numbers each time your program is run.
Also, does it have any tendencies
towards odd or even numbers or
something like that?
Depends on the exact method used. There's one popular implementation of rand() that alternates between odd and even numbers. So avoid writing code like rand() % 2 that depends on the lowest bit being random.
This question already has answers here:
Probability of SHA1 collisions
(3 answers)
Closed 6 years ago.
Given two different strings S1 and S2 (S1 != S2) is it possible that:
SHA1(S1) == SHA1(S2)
is True?
If yes - with what probability?
If not - why not?
Is there a upper bound on the length of a input string, for which the probability of getting duplicates is 0? OR is the calculation of SHA1 (hence probability of duplicates) independent of the length of the string?
The goal I am trying to achieve is to hash some sensitive ID string (possibly joined together with some other fields like parent ID), so that I can use the hash value as an ID instead (for example in the database).
Example:
Resource ID: X123
Parent ID: P123
I don't want to expose the nature of my resource identifies to allow client to see "X123-P123".
Instead I want to create a new column hash("X123-P123"), let's say it's AAAZZZ. Then the client can request resource with id AAAZZZ and not know about my internal id's etc.
What you describe is called a collision. Collisions necessarily exist, since SHA-1 accepts many more distinct messages as input that it can produce distinct outputs (SHA-1 may eat any string of bits up to 2^64 bits, but outputs only 160 bits; thus, at least one output value must pop up several times). This observation is valid for any function with an output smaller than its input, regardless of whether the function is a "good" hash function or not.
Assuming that SHA-1 behaves like a "random oracle" (a conceptual object which basically returns random values, with the sole restriction that once it has returned output v on input m, it must always thereafter return v on input m), then the probability of collision, for any two distinct strings S1 and S2, should be 2^(-160). Still under the assumption of SHA-1 behaving like a random oracle, if you collect many input strings, then you shall begin to observe collisions after having collected about 2^80 such strings.
(That's 2^80 and not 2^160 because, with 2^80 strings you can make about 2^159 pairs of strings. This is often called the "birthday paradox" because it comes as a surprise to most people when applied to collisions on birthdays. See the Wikipedia page on the subject.)
Now we strongly suspect that SHA-1 does not really behave like a random oracle, because the birthday-paradox approach is the optimal collision searching algorithm for a random oracle. Yet there is a published attack which should find a collision in about 2^63 steps, hence 2^17 = 131072 times faster than the birthday-paradox algorithm. Such an attack should not be doable on a true random oracle. Mind you, this attack has not been actually completed, it remains theoretical (some people tried but apparently could not find enough CPU power)(Update: as of early 2017, somebody did compute a SHA-1 collision with the above-mentioned method, and it worked exactly as predicted). Yet, the theory looks sound and it really seems that SHA-1 is not a random oracle. Correspondingly, as for the probability of collision, well, all bets are off.
As for your third question: for a function with a n-bit output, then there necessarily are collisions if you can input more than 2^n distinct messages, i.e. if the maximum input message length is greater than n. With a bound m lower than n, the answer is not as easy. If the function behaves as a random oracle, then the probability of the existence of a collision lowers with m, and not linearly, rather with a steep cutoff around m=n/2. This is the same analysis than the birthday paradox. With SHA-1, this means that if m < 80 then chances are that there is no collision, while m > 80 makes the existence of at least one collision very probable (with m > 160 this becomes a certainty).
Note that there is a difference between "there exists a collision" and "you find a collision". Even when a collision must exist, you still have your 2^(-160) probability every time you try. What the previous paragraph means is that such a probability is rather meaningless if you cannot (conceptually) try 2^160 pairs of strings, e.g. because you restrict yourself to strings of less than 80 bits.
Yes it is possible because of the pigeon hole principle.
Most hashes (also sha1) have a fixed output length, while the input is of arbitrary size. So if you try long enough, you can find them.
However, cryptographic hash functions (like the sha-family, the md-family, etc) are designed to minimize such collisions. The best attack known takes 2^63 attempts to find a collision, so the chance is 2^(-63) which is 0 in practice.
git uses SHA1 hashes as IDs and there are still no known SHA1 collisions in 2014. Obviously, the SHA1 algorithm is magic. I think it's a good bet that collisions don't exist for strings of your length, as they would have been discovered by now. However, if you don't trust magic and are not a betting man, you could generate random strings and associate them with your IDs in your DB. But if you do use SHA1 hashes and become the first to discover a collision, you can just change your system to use random strings at that time, retaining the SHA1 hashes as the "random" strings for legacy IDs.
A collision is almost always possible in a hashing function. SHA1, to date, has been pretty secure in generating unpredictable collisions. The danger is when collisions can be predicted, it's not necessary to know the original hash input to generate the same hash output.
For example, attacks against MD5 have been made against SSL server certificate signing last year, as exampled on the Security Now podcast episode 179. This allowed sophisticated attackers to generate a fake SSL server cert for a rogue web site and appear to be the reaol thing. For this reason, it is highly recommended to avoid purchasing MD5-signed certs.
What you are talking about is called a collision. Here is an article about SHA1 collisions:
http://www.rsa.com/rsalabs/node.asp?id=2927
Edit: So another answerer beat me to mentioning the pigeon hole principle LOL, but to clarify this is why it's called the pigeon hole principle, because if you have some holes cut out for carrier pigeons to nest in, but you have more pigeons than holes, then some of the pigeons(an input value) must share a hole(the output value).
I have read through various papers on the 'Balls and Bins' problem and it seems that if a hash function is working right (ie. it is effectively a random distribution) then the following should/must be true if I hash n values into a hash table with n slots (or bins):
Probability that a bin is empty, for large n is 1/e.
Expected number of empty bins is n/e.
Probability that a bin has k balls is <= 1/ek! (corrected).
Probability that a bin has at least k collisions is <= ((e/k)**k)/e (corrected).
These look easy to check. But the max-load test (the maximum number of collisions with high probability) is usually stated vaguely.
Most texts state that the maximum number of collisions in any bin is O( ln(n) / ln(ln(n)) ).
Some say it is 3*ln(n) / ln(ln(n)). Other papers mix ln and log - usually without defining them, or state that log is log base e and then use ln elsewhere.
Is ln the log to base e or 2 and is this max-load formula right and how big should n be to run a test?
This lecture seems to cover it best, but I am no mathematician.
http://pages.cs.wisc.edu/~shuchi/courses/787-F07/scribe-notes/lecture07.pdf
BTW, with high probability seems to mean 1 - 1/n.
That is a fascinating paper/lecture-- makes me wish I had taken some formal algorithms class.
I'm going to take a stab at some answers here, based on what I've just read from that, and feel free to vote me down. I'd appreciate a correction, though, rather than just a downvote :) I'm also going to use n and N interchangeably here, which is a big no-no in some circles, but since I'm just copy-pasting your formulae, I hope you'll forgive me.
First, the base of the logs. These numbers are given as big-O notation, not as absolute formulae. That means that you're looking for something 'on the order of ln(n) / ln(ln(n))', not with an expectation of an absolute answer, but more that as n gets bigger, the relationship of n to the maximum number of collisions should follow that formula. The details of the actual curve you can graph will vary by implementation (and I don't know enough about the practical implementations to tell you what's a 'good' curve, except that it should follow that big-O relationship). Those two formulae that you posted are actually equivalent in big-O notation. The 3 in the second formula is just a constant, and is related to a particular implementation. A less efficient implementation would have a bigger constant.
With that in mind, I would run empirical tests, because I'm a biologist at heart and I was trained to avoid hard-and-fast proofs as indications of how the world actually works. Start with N as some number, say 100, and find the bin with the largest number of collisions in it. That's your max-load for that run. Now, your examples should be as close as possible to what you expect actual users to use, so maybe you want to randomly pull words from a dictionary or something similar as your input.
Run that test many times, at least 30 or 40. Since you're using random numbers, you'll need to satisfy yourself that the average max-load you're getting is close to the theoretical 'expectation' of your algorithm. Expectation is just the average, but you'll still need to find it, and the tighter your std dev/std err about that average, the more you can say that your empirical average matches the theoretical expectation. One run is not enough, because a second run will (most likely) give a different answer.
Then, increase N, to say, 1000, 10000, etc. Increase it logarithmically, because your formula is logarithmic. As your N increases, your max-load should increase on the order of ln(n) / ln(ln(n)). If it increases at a rate of 3*ln(n) / ln(ln(n)), that means that you're following the theory that they put forth in that lecture.
This kind of empirical test will also show you where your approach breaks down. It may be that your algorithm works well for N < 10 million (or some other number), but above that, it starts to collapse. Why could that be? Maybe you have some limitation to 32 bits in your code without realizing it (ie, using a 'float' instead of a 'double'), or some other implementation detail. These kinds of details let you know where your code will work well in practice, and then as your practical needs change, you can modify your algorithm. Maybe making the algorithm work for very large datasets makes it very inefficient for very small ones, or vice versa, so pinpointing that tradeoff will help you further characterize how you could adapt your algorithm to particular situations. Always a useful skill to have.
EDIT: a proof of why the base of the log function doesn't matter with big-O notation:
log N = log_10 (N) = log_b (N)/log_b (10)= (1/log_b(10)) * log_b(N)
1/log_b(10) is a constant, and in big-O notation, constants are ignored. Base changes are free, which is why you're encountering such variation in the papers.
Here is a rough start to the solution of this problem involving uniform distributions and maximum load.
Instead of bins and balls or urns or boxes or buckets or m and n, people (p) and doors (d) will be used as designations.
There is an exact expected value for each of the doors given a certain number of people. For example, with 5 people and 5 doors, the expected maximum door is exactly 1.2864 {(1429-625) / 625} above the mean (p/d) and the minimum door is exactly -0.9616 {(24-625) / 625} below the mean. The absolute value of the highest door's distance from the mean is a little larger than the smallest door's because all of the people could go through one door, but no less than zero can go through one of the doors. With large numbers of people (p/d > 3000), the difference between the absolute value of the highest door's distance from the mean and the lowest door's becomes negligible.
For an odd number of doors, the center door is essentially zero and is not scalable, but all of the other doors are scalable from certain values representing p=d. These rounded values for d=5 are:
-1.163 -0.495 0* 0.495 1.163
* slowly approaching zero from -0.12
From these values, you can compute the expected number of people for any count of people going through each of the 5 doors, including the maximum door. Except for the middle ordered door, the difference from the mean is scalable by sqrt(p/d).
So, for p=50,000 and d=5:
Expected number of people going through the maximum door, which could be any of the 5 doors, = 1.163 * sqrt(p/d) + p/d.
= 1.163 * sqrt(10,000) + 10,000 = 10,116.3
For p/d < 3,000, the result from this equation must be slightly increased.
With more people, the middle door slowly becomes closer and closer to zero from -0.11968 at p=100 and d=5. It can always be rounded up to zero and like the other 4 doors has quite a variance.
The values for 6 doors are:
-1.272 -0.643 -0.202 0.202 0.643 1.272
For 1000 doors, the approximate values are:
-3.25, -2.95, -2.79 … 2.79, 2.95, 3.25
For any d and p, there is an exact expected value for each of the ordered doors. Hopefully, a good approximation (with a relative error < 1%) exists. Some professor or mathematician somewhere must know.
For testing uniform distribution, you will need a number of averaged ordered sessions (750-1000 works well) rather than a greater number of people. No matter what, the variances between valid sessions are great. That's the nature of randomness. Collisions are unavoidable. *
The expected values for 5 and 6 doors were obtained by sheer brute force computation using 640 bit integers and averaging the convergence of the absolute values of corresponding opposite doors.
For d=5 and p=170:
-6.63901 -2.95905 -0.119342 2.81054 6.90686
(27.36099 31.04095 33.880658 36.81054 40.90686)
For d=6 and p=108:
-5.19024 -2.7711 -0.973979 0.734434 2.66716 5.53372
(12.80976 15.2289 17.026021 18.734434 20.66716 23.53372)
I hope that you may evenly distribute your data.
It's almost guaranteed that all of George Foreman's sons or some similar situation will fight against your hash function. And proper contingent planning is the work of all good programmers.
After some more research and trial-and-error I think I can provide something part way to to an answer.
To start off, ln and log seem to refer to log base-e if you look into the maths behind the theory. But as mmr indicated, for the O(...) estimates, it doesn't matter.
max-load can be defined for any probability you like. The typical formula used is
1-1/n**c
Most papers on the topic use
1-1/n
An example might be easiest.
Say you have a hash table of 1000 slots and you want to hash 1000 things. Say you also want to know the max-load with a probability of 1-1/1000 or 0.999.
The max-load is the maximum number of hash values that end up being the same - ie. collisions (assuming that your hash function is good).
Using the formula for the probability of getting exactly k identical hash values
Pr[ exactly k ] = ((e/k)**k)/e
then by accumulating the probability of exactly 0..k items until the total equals or exceeds 0.999 tells you that k is the max-load.
eg.
Pr[0] = 0.37
Pr[1] = 0.37
Pr[2] = 0.18
Pr[3] = 0.061
Pr[4] = 0.015
Pr[5] = 0.003 // here, the cumulative total is 0.999
Pr[6] = 0.0005
Pr[7] = 0.00007
So, in this case, the max-load is 5.
So if my hash function is working well on my set of data then I should expect the maxmium number of identical hash values (or collisions) to be 5.
If it isn't then this could be due to the following reasons:
Your data has small values (like short strings) that hash to the same value. Any hash of a single ASCII character will pick 1 of 128 hash values (there are ways around this. For example you could use multiple hash functions, but slows down hashing and I don't know much about this).
Your hash function doesn't work well with your data - try it with random data.
Your hash function doesn't work well.
The other tests I mentioned in my question also are helpful to see that your hash function is running as expected.
Incidentally, my hash function worked nicely - except on short (1..4 character) strings.
I also implemented a simple split-table version which places the hash value into the least used slot from a choice of 2 locations. This more than halves the number of collisions and means that adding and searching the hash table is a little slower.
I hope this helps.