Universal hashing

In mathematics and computing, universal hashing (in a randomized algorithm or data structure) refers to selecting a hash function at random from a family of hash functions with a certain mathematical property (see definition below). This guarantees a low number of collisions in expectation, even if the data is chosen by an adversary. Many universal families are known (for hashing integers, vectors, strings), and their evaluation is often very efficient. Universal hashing has numerous uses in computer science, for example in implementations of hash tables, randomized algorithms, and cryptography.


Assume we want to map keys from some universe   into   bins (labelled  ). The algorithm will have to handle some data set   of   keys, which is not known in advance. Usually, the goal of hashing is to obtain a low number of collisions (keys from   that land in the same bin). A deterministic hash function cannot offer any guarantee in an adversarial setting if the size of   is greater than  , since the adversary may choose   to be precisely the preimage of a bin. This means that all data keys land in the same bin, making hashing useless. Furthermore, a deterministic hash function does not allow for rehashing: sometimes the input data turns out to be bad for the hash function (e.g. there are too many collisions), so one would like to change the hash function.

The solution to these problems is to pick a function randomly from a family of hash functions. A family of functions   is called a universal family if,  .

In other words, any two keys of the universe collide with probability at most   when the hash function   is drawn randomly from  . This is exactly the probability of collision we would expect if the hash function assigned truly random hash codes to every key. Sometimes, the definition is relaxed to allow collision probability  . This concept was introduced by Carter and Wegman[1] in 1977, and has found numerous applications in computer science (see, for example [2]). If we have an upper bound of   on the collision probability, we say that we have  -almost universality.

Many, but not all, universal families have the following stronger uniform difference property:

 , when   is drawn randomly from the family  , the difference   is uniformly distributed in  .

Note that the definition of universality is only concerned with whether  , which counts collisions. The uniform difference property is stronger.

(Similarly, a universal family can be XOR universal if  , the value   is uniformly distributed in   where   is the bitwise exclusive or operation. This is only possible if   is a power of two.)

An even stronger condition is pairwise independence: we have this property when   we have the probability that   will hash to any pair of hash values   is as if they were perfectly random:  . Pairwise independence is sometimes called strong universality.

Another property is uniformity. We say that a family is uniform if all hash values are equally likely:   for any hash value  . Universality does not imply uniformity. However, strong universality does imply uniformity.

Given a family with the uniform distance property, one can produce a pairwise independent or strongly universal hash family by adding a uniformly distributed random constant with values in   to the hash functions. (Similarly, if   is a power of two, we can achieve pairwise independence from an XOR universal hash family by doing an exclusive or with a uniformly distributed random constant.) Since a shift by a constant is sometimes irrelevant in applications (e.g. hash tables), a careful distinction between the uniform distance property and pairwise independent is sometimes not made.[3]

For some applications (such as hash tables), it is important for the least significant bits of the hash values to be also universal. When a family is strongly universal, this is guaranteed: if   is a strongly universal family with  , then the family made of the functions   for all   is also strongly universal for  . Unfortunately, the same is not true of (merely) universal families. For example, the family made of the identity function   is clearly universal, but the family made of the function   fails to be universal.

UMAC and Poly1305-AES and several other message authentication code algorithms are based on universal hashing.[4][5] In such applications, the software chooses a new hash function for every message, based on a unique nonce for that message.

Several hash table implementations are based on universal hashing. In such applications, typically the software chooses a new hash function only after it notices that "too many" keys have collided; until then, the same hash function continues to be used over and over. (Some collision resolution schemes, such as dynamic perfect hashing, pick a new hash function every time there is a collision. Other collision resolution schemes, such as cuckoo hashing and 2-choice hashing, allow a number of collisions before picking a new hash function). A survey of fastest known universal and strongly universal hash functions for integers, vectors, and strings is found in.[6]

Mathematical guaranteesEdit

For any fixed set   of   keys, using a universal family guarantees the following properties.

  1. For any fixed   in  , the expected number of keys in the bin   is  . When implementing hash tables by chaining, this number is proportional to the expected running time of an operation involving the key   (for example a query, insertion or deletion).
  2. The expected number of pairs of keys   in   with   that collide ( ) is bounded above by  , which is of order  . When the number of bins,   is chosen linear in   (i.e., is determined by a function in  ), the expected number of collisions is  . When hashing into   bins, there are no collisions at all with probability at least a half.
  3. The expected number of keys in bins with at least   keys in them is bounded above by  .[7] Thus, if the capacity of each bin is capped to three times the average size ( ), the total number of keys in overflowing bins is at most  . This only holds with a hash family whose collision probability is bounded above by  . If a weaker definition is used, bounding it by  , this result is no longer true.[7]

As the above guarantees hold for any fixed set  , they hold if the data set is chosen by an adversary. However, the adversary has to make this choice before (or independent of) the algorithm's random choice of a hash function. If the adversary can observe the random choice of the algorithm, randomness serves no purpose, and the situation is the same as deterministic hashing.

The second and third guarantee are typically used in conjunction with rehashing. For instance, a randomized algorithm may be prepared to handle some   number of collisions. If it observes too many collisions, it chooses another random   from the family and repeats. Universality guarantees that the number of repetitions is a geometric random variable.


Since any computer data can be represented as one or more machine words, one generally needs hash functions for three types of domains: machine words ("integers"); fixed-length vectors of machine words; and variable-length vectors ("strings").

Hashing integersEdit

This section refers to the case of hashing integers that fit in machines words; thus, operations like multiplication, addition, division, etc. are cheap machine-level instructions. Let the universe to be hashed be  .

The original proposal of Carter and Wegman[1] was to pick a prime   and define


where   are randomly chosen integers modulo   with  . (This is a single iteration of a linear congruential generator.)

To see that   is a universal family, note that   only holds when


for some integer   between   and  . Since  , if   their difference   is nonzero and has an inverse modulo  . Solving for   yields


There are   possible choices for   (since   is excluded) and, varying   in the allowed range,   possible non-zero values for the right hand side. Thus the collision probability is


Another way to see   is a universal family is via the notion of statistical distance. Write the difference   as


Since   is nonzero and   is uniformly distributed in  , it follows that   modulo   is also uniformly distributed in  . The distribution of   is thus almost uniform, up to a difference in probability of   between the samples. As a result, the statistical distance to a uniform family is  , which becomes negligible when  .

The family of simpler hash functions


is only approximately universal:   for all  .[1] Moreover, this analysis is nearly tight; Carter and Wegman [1] show that   whenever  .

Avoiding modular arithmeticEdit

The state of the art for hashing integers is the multiply-shift scheme described by Dietzfelbinger et al. in 1997.[8] By avoiding modular arithmetic, this method is much easier to implement and also runs significantly faster in practice (usually by at least a factor of four[9]). The scheme assumes the number of bins is a power of two,  . Let   be the number of bits in a machine word. Then the hash functions are parametrised over odd positive integers   (that fit in a word of   bits). To evaluate  , multiply   by   modulo   and then keep the high order   bits as the hash code. In mathematical notation, this is


and it can be implemented in C-like programming languages by

  (size_t) (a*x) >> (w-M)

This scheme does not satisfy the uniform difference property and is only  -almost-universal; for any  ,  .

To understand the behavior of the hash function, notice that, if   and   have the same highest-order 'M' bits, then   has either all 1's or all 0's as its highest order M bits (depending on whether   or   is larger). Assume that the least significant set bit of   appears on position  . Since   is a random odd integer and odd integers have inverses in the ring  , it follows that   will be uniformly distributed among  -bit integers with the least significant set bit on position  . The probability that these bits are all 0's or all 1's is therefore at most  . On the other hand, if  , then higher-order M bits of   contain both 0's and 1's, so it is certain that  . Finally, if   then bit   of   is 1 and   if and only if bits   are also 1, which happens with probability  .

This analysis is tight, as can be shown with the example   and  . To obtain a truly 'universal' hash function, one can use the multiply-add-shift scheme


which can be implemented in C-like programming languages by

  (size_t) (a*x+b) >> (w-M)

where   is a random odd positive integer with   and   is a random non-negative integer with  . With these choices of   and  ,   for all  .[10] This differs slightly but importantly from the mistranslation in the English paper.[11]

Hashing vectorsEdit

This section is concerned with hashing a fixed-length vector of machine words. Interpret the input as a vector   of   machine words (integers of   bits each). If   is a universal family with the uniform difference property, the following family (dating back to Carter and Wegman[1]) also has the uniform difference property (and hence is universal):

 , where each   is chosen independently at random.

If   is a power of two, one may replace summation by exclusive or.[12]

In practice, if double-precision arithmetic is available, this is instantiated with the multiply-shift hash family of hash functions.[13] Initialize the hash function with a vector   of random odd integers on   bits each. Then if the number of bins is   for  :


It is possible to halve the number of multiplications, which roughly translates to a two-fold speed-up in practice.[12] Initialize the hash function with a vector   of random odd integers on   bits each. The following hash family is universal:[14]


If double-precision operations are not available, one can interpret the input as a vector of half-words ( -bit integers). The algorithm will then use   multiplications, where   was the number of half-words in the vector. Thus, the algorithm runs at a "rate" of one multiplication per word of input.

The same scheme can also be used for hashing integers, by interpreting their bits as vectors of bytes. In this variant, the vector technique is known as tabulation hashing and it provides a practical alternative to multiplication-based universal hashing schemes.[15]

Strong universality at high speed is also possible.[16] Initialize the hash function with a vector   of random integers on   bits. Compute


The result is strongly universal on   bits. Experimentally, it was found to run at 0.2 CPU cycle per byte on recent Intel processors for  .

Hashing stringsEdit

This refers to hashing a variable-sized vector of machine words. If the length of the string can be bounded by a small number, it is best to use the vector solution from above (conceptually padding the vector with zeros up to the upper bound). The space required is the maximal length of the string, but the time to evaluate   is just the length of  . As long as zeroes are forbidden in the string, the zero-padding can be ignored when evaluating the hash function without affecting universality.[12] Note that if zeroes are allowed in the string, then it might be best to append a fictitious non-zero (e.g., 1) character to all strings prior to padding: this will ensure that universality is not affected.[16]

Now assume we want to hash  , where a good bound on   is not known a priori. A universal family proposed by [13] treats the string   as the coefficients of a polynomial modulo a large prime. If  , let   be a prime and define:

 , where   is uniformly random and   is chosen randomly from a universal family mapping integer domain  .

Using properties of modular arithmetic, above can be computed without producing large numbers for large strings as follows:[17]

uint hash(String x, int a, int p)
	for (uint i=0 ; i < x.length ; ++i)
		h = ((h*a) + x[i]) mod p
	return h

This Rabin-Karp rolling hash is based on a linear congruential generator.[18] Above algorithm is also known as Multiplicative hash function.[19] In practice, the mod operator and the parameter p can be avoided altogether by simply allowing integer to overflow because it is equivalent to mod (Max-Int-Value + 1) in many programming languages. Below table shows values chosen to initialize h and a for some of the popular implementations.

Implementation INITIAL_VALUE a
Bernstein's hash function djb2[20] 5381 33
STLPort 4.6.2 0 5
Kernighan and Ritchie's hash function[21] 0 31
java.lang.String.hashCode()[22] 0 31

Consider two strings   and let   be length of the longer one; for the analysis, the shorter string is conceptually padded with zeros up to length  . A collision before applying   implies that   is a root of the polynomial with coefficients  . This polynomial has at most   roots modulo  , so the collision probability is at most  . The probability of collision through the random   brings the total collision probability to  . Thus, if the prime   is sufficiently large compared to the length of strings hashed, the family is very close to universal (in statistical distance).

Other universal families of hash functions used to hash unknown-length strings to fixed-length hash values include the Rabin fingerprint and the Buzhash.

Avoiding modular arithmeticEdit

To mitigate the computational penalty of modular arithmetic, three tricks are used in practice:[12]

  1. One chooses the prime   to be close to a power of two, such as a Mersenne prime. This allows arithmetic modulo   to be implemented without division (using faster operations like addition and shifts). For instance, on modern architectures one can work with  , while  's are 32-bit values.
  2. One can apply vector hashing to blocks. For instance, one applies vector hashing to each 16-word block of the string, and applies string hashing to the   results. Since the slower string hashing is applied on a substantially smaller vector, this will essentially be as fast as vector hashing.
  3. One chooses a power-of-two as the divisor, allowing arithmetic modulo   to be implemented without division (using faster operations of bit masking). The NH hash-function family takes this approach.

See alsoEdit


  1. ^ a b c d e Carter, Larry; Wegman, Mark N. (1979). "Universal Classes of Hash Functions". Journal of Computer and System Sciences. 18 (2): 143–154. doi:10.1016/0022-0000(79)90044-8. Conference version in STOC'77.
  2. ^ Miltersen, Peter Bro. "Universal Hashing" (PDF). Archived from the original (PDF) on 24 May 2011. Retrieved 24 June 2009.
  3. ^ Motwani, Rajeev; Raghavan, Prabhakar (1995). Randomized Algorithms. Cambridge University Press. p. 221. ISBN 0-521-47465-5.
  4. ^ David Wagner, ed. "Advances in Cryptology - CRYPTO 2008". p. 145.
  5. ^ Jean-Philippe Aumasson, Willi Meier, Raphael Phan, Luca Henzen. "The Hash Function BLAKE". 2014. p. 10.
  6. ^ Thorup, Mikkel (2015). "High Speed Hashing for Integers and Strings". arXiv:1504.06804 [cs.DS].
  7. ^ a b Baran, Ilya; Demaine, Erik D.; Pătraşcu, Mihai (2008). "Subquadratic Algorithms for 3SUM" (PDF). Algorithmica. 50 (4): 584–596. doi:10.1007/s00453-007-9036-3.
  8. ^ Dietzfelbinger, Martin; Hagerup, Torben; Katajainen, Jyrki; Penttonen, Martti (1997). "A Reliable Randomized Algorithm for the Closest-Pair Problem" (Postscript). Journal of Algorithms. 25 (1): 19–51. doi:10.1006/jagm.1997.0873. Retrieved 10 February 2011.
  9. ^ Thorup, Mikkel. "Text-book algorithms at SODA".
  10. ^ Woelfel, Philipp (2003). Über die Komplexität der Multiplikation in eingeschränkten Branchingprogrammmodellen (PDF) (Ph.D.). Universität Dortmund. Retrieved 18 September 2012.
  11. ^ Woelfel, Philipp (1999). Efficient Strongly Universal and Optimally Universal Hashing. Mathematical Foundations of Computer Science 1999. LNCS. 1672. pp. 262–272. doi:10.1007/3-540-48340-3_24.
  12. ^ a b c d Thorup, Mikkel (2009). String hashing for linear probing. Proc. 20th ACM-SIAM Symposium on Discrete Algorithms (SODA). pp. 655–664. CiteSeerX doi:10.1137/1.9781611973068.72., section 5.3
  13. ^ a b Dietzfelbinger, Martin; Gil, Joseph; Matias, Yossi; Pippenger, Nicholas (1992). Polynomial Hash Functions Are Reliable (Extended Abstract). Proc. 19th International Colloquium on Automata, Languages and Programming (ICALP). pp. 235–246.
  14. ^ Black, J.; Halevi, S.; Krawczyk, H.; Krovetz, T. (1999). UMAC: Fast and Secure Message Authentication (PDF). Advances in Cryptology (CRYPTO '99)., Equation 1
  15. ^ Pătraşcu, Mihai; Thorup, Mikkel (2011). The power of simple tabulation hashing. Proceedings of the 43rd annual ACM Symposium on Theory of Computing (STOC '11). pp. 1–10. arXiv:1011.5200. doi:10.1145/1993636.1993638.
  16. ^ a b Kaser, Owen; Lemire, Daniel (2013). "Strongly universal string hashing is fast". Computer Journal. Oxford University Press. 57 (11): 1624–1638. arXiv:1202.4961. doi:10.1093/comjnl/bxt070.
  17. ^ "Hebrew University Course Slides" (PDF).
  18. ^ Robert Uzgalis. "Library Hash Functions". 1996.
  19. ^ Kankowsk, Peter. "Hash functions: An empirical comparison".
  20. ^ Yigit, Ozan. "String hash functions".
  21. ^ Kernighan; Ritchie (1988). "6". The C Programming Language (2nd ed.). pp. 118. ISBN 0-13-110362-8.CS1 maint: multiple names: authors list (link)
  22. ^ "String (Java Platform SE 6)". docs.oracle.com. Retrieved 2015-06-10.

Further readingEdit

External linksEdit