Park-Miller LCG

How Park-Miller LCGs work

A Park-Miller linear congruential generator is stateful, deterministic, and simple by design. Each new value is derived from the previous state through a fixed modular multiplication, which makes the entire sequence reproducible from a seed. That simplicity is useful for simulation, testing, and teaching, but it also means the generator does not create unpredictability on its own.

The key property to understand is that the output is only as strong as the seed and the algorithm. If both are known, the sequence can be recreated exactly. For that reason, a Park-Miller LCG should be treated as a predictable pseudo-random sequence generator, not as a source of security-grade randomness.

Why deterministic output matters

Determinism is the defining feature of the Park-Miller LCG family. Given the same seed, it produces the same sequence every time, which is helpful when you need repeatable results across runs, systems, or test environments.

That same property becomes a limitation in security-sensitive settings. Predictable sequences can be inferred, replayed, or matched against observed outputs when the seed is weak, reused, or exposed. In other words, the mechanism is mathematically consistent, but it does not provide meaningful resistance against prediction.

Typical uses and practical limits

Park-Miller LCGs have long been used where speed, portability, and reproducibility matter more than statistical quality. They are often acceptable for basic sampling, legacy software, or controlled experiments where repeatability is the goal.

They are a poor fit for token generation, secrets, session identifiers, keys, or any workflow that depends on unpredictability. In those cases, a modern cryptographically secure generator is required because the business need is not just variation, but resistance to prediction and reconstruction.

How to think about seed management

The seed is the control point for the entire sequence. If the seed is chosen poorly, reused across contexts, or derived from an observable value, the resulting sequence can become guessable even if the generator itself is implemented correctly.

For practitioners, the main question is not whether the generator is functioning as specified, but whether the predictability of that specification is acceptable for the intended use. A secure design should separate reproducible pseudo-randomness for non-sensitive tasks from cryptographic randomness for anything that depends on secrecy or trust.

Risk and Threat Considerations

Predictable pseudo-random output creates exposure when a Park-Miller LCG is used in places that depend on entropy. An attacker who learns the algorithm and seed can reconstruct future values, and weak seed handling can make that feasible even without direct access to the generator.

Failure mechanism: The generator is deterministic, so weak, reused, or exposed seeds let an adversary reproduce the same sequence and anticipate values that were assumed to be random.

Impact: Prediction can undermine session tokens, test data isolation, challenge values, or any workflow that mistakenly treats the sequence as a secret or as a source of freshness.

Practitioner Guidance

Why practitioners should care: The main decision is whether reproducibility or unpredictability is the real requirement. Park-Miller LCGs are fine when consistency is the goal, but they should not be used where an observer gaining knowledge of one output could use it to infer the next.

What to watch for: Review any use of an LCG in code that also handles authentication, access tokens, salts, identifiers, or security-sensitive sampling. If the same generator is serving both testing and production needs, the security boundary is usually too thin.

Practitioner takeaway: Use Park-Miller-style generators only where deterministic repeatability is a feature, and reserve cryptographically secure randomness for anything that must remain unpredictable.