When a protection mechanism relies on random values to restrict access to a sensitive resource, such as a session ID or a seed for generating a cryptographic key, then the resource being protected could be accessed by guessing the ID or key.

If software relies on unique, unguessable IDs to identify a resource, an attacker might be able to guess an ID for a resource that is owned by another user. The attacker could then read the resource, or pre-create a resource with the same ID to prevent the legitimate program from properly sending the resource to the intended user. For example, a product might maintain session information in a file whose name is based on a username. An attacker could pre-create this file for a victim user, then set the permissions so that the application cannot generate the session for the victim, preventing the victim from using the application.

When an authorization or authentication mechanism relies on random values to restrict access to restricted functionality, such as a session ID or a seed for generating a cryptographic key, then an attacker may access the restricted functionality by guessing the ID or key.

Use a well-vetted algorithm that is currently considered to be strong by experts in the field, and select well-tested implementations with adequate length seeds.

In general, if a pseudo-random number generator is not advertised as being cryptographically secure, then it is probably a statistical PRNG and should not be used in security-sensitive contexts.

Pseudo-random number generators can produce predictable numbers if the generator is known and the seed can be guessed. A 256-bit seed is a good starting point for producing a "random enough" number.

Consider a PRNG that re-seeds itself as needed from high quality pseudo-random output sources, such as hardware devices.

Use automated static analysis tools that target this type of weakness. Many modern techniques use data flow analysis to minimize the number of false positives. This is not a perfect solution, since 100% accuracy and coverage are not feasible.

Perform FIPS 140-2 tests on data to catch obvious entropy problems.

Use tools and techniques that require manual (human) analysis, such as penetration testing, threat modeling, and interactive tools that allow the tester to record and modify an active session. These may be more effective than strictly automated techniques. This is especially the case with weaknesses that are related to design and business rules.

Use monitoring tools that examine the software's process as it interacts with the operating system and the network. This technique is useful in cases when source code is unavailable, if the software was not developed by you, or if you want to verify that the build phase did not introduce any new weaknesses. Examples include debuggers that directly attach to the running process; system-call tracing utilities such as truss (Solaris) and strace (Linux); system activity monitors such as FileMon, RegMon, Process Monitor, and other Sysinternals utilities (Windows); and sniffers and protocol analyzers that monitor network traffic.

Attach the monitor to the process and look for library functions that indicate when randomness is being used. Run the process multiple times to see if the seed changes. Look for accesses of devices or equivalent resources that are commonly used for strong (or weak) randomness, such as /dev/urandom on Linux. Look for library or system calls that access predictable information such as process IDs and system time.