| Globally unique identifier generator.
In the multitude of java GUID generators, I found none that guaranteed
randomness. GUIDs are guaranteed to be globally unique by using ethernet
MACs, IP addresses, time elements, and sequential numbers. GUIDs are not
expected to be random and most often are easy/possible to guess given a
sample from a given generator. SQL Server, for example generates GUID that
are unique but sequencial within a given instance.
GUIDs can be used as security devices to hide things such as files within a
filesystem where listings are unavailable (e.g. files that are served up from
a Web server with indexing turned off). This may be desireable in cases where
standard authentication is not appropriate. In this scenario, the RandomGuids
are used as directories. Another example is the use of GUIDs for primary keys
in a database where you want to ensure that the keys are secret. Random GUIDs
can then be used in a URL to prevent hackers (or users) from accessing
records by guessing or simply by incrementing sequential numbers.
There are many other possiblities of using GUIDs in the realm of security and
encryption where the element of randomness is important. This class was
written for these purposes but can also be used as a general purpose GUID
generator as well.
RandomGuid generates truly random GUIDs by using the system's IP address
(name/IP), system time in milliseconds (as an integer), and a very large
random number joined together in a single String that is passed through an
MD5 hash. The IP address and system time make the MD5 seed globally unique
and the random number guarantees that the generated GUIDs will have no
discernable pattern and cannot be guessed given any number of previously
generated GUIDs. It is generally not possible to access the seed information
(IP, time, random number) from the resulting GUIDs as the MD5 hash algorithm
provides one way encryption.
Security of RandomGuid: RandomGuid can be called one of two ways --
with the basic java Random number generator or a cryptographically strong
random generator (SecureRandom). The choice is offered because the secure
random generator takes about 3.5 times longer to generate its random numbers
and this performance hit may not be worth the added security especially
considering the basic generator is seeded with a cryptographically strong
random seed.
Seeding the basic generator in this way effectively decouples the random
numbers from the time component making it virtually impossible to predict the
random number component even if one had absolute knowledge of the System
time. Thanks to Ashutosh Narhari for the suggestion of using the static
method to prime the basic random generator.
Using the secure random option, this class complies with the statistical
random number generator tests specified in FIPS 140-2, Security Requirements
for Cryptographic Modules, secition 4.9.1.
I converted all the pieces of the seed to a String before handing it over to
the MD5 hash so that you could print it out to make sure it contains the data
you expect to see and to give a nice warm fuzzy. If you need better
performance, you may want to stick to byte[] arrays.
I believe that it is important that the algorithm for generating random GUIDs
be open for inspection and modification. This class is free for all uses.
version:
1.2.1 11/05/02
author:
Marc A. Mnich | 