Randar: A Minecraft exploit that uses LLL lattice reduction to crack server RNG

Original Article ↗ Hacker News Discussion ↗

A newly documented Minecraft exploit shows how attackers used LLL lattice reduction and weaknesses in the game’s shared pseudo‑random number generator to infer server RNG state, locate valuable player bases, and trigger precise TNT strikes on the anarchy server 2b2t. Commenters use it as a case study in why non‑cryptographic PRNGs and shared RNG state are unsafe for anything security‑sensitive, contrasting them with modern CSPRNGs and strategies like isolating multiple RNG instances. The thread also touches on broader game‑metagaming culture, where bugs, exploits, and reverse engineering effectively become an extra competitive layer on top of the nominal game rules.

RNG Vulnerability & Design Lessons

  • Thread treats the exploit as a textbook case of why sharing RNG state across subsystems is dangerous.
  • Several comments note this is not a flaw in the PRNG algorithm itself, but in using a non-cryptographic PRNG where security properties are implicitly needed.
  • Suggested fixes split between “just use a CSPRNG everywhere” and “use many separate, non‑crypto PRNGs with isolated state so compromising one reveals nothing about others.”
  • The attack is framed as a State Compromise Extension Attack: once state is inferred from observations, future outputs become predictable.
  • Some argue libraries should default to CSPRNGs in 2024, much like modern TLS defaults.

CSPRNG vs Game PRNG Performance

  • One side claims CSPRNGs would have made early Minecraft too slow (chunk loading, random placement), favoring ultra‑cheap PRNGs.
  • Others counter that modern CSPRNGs (e.g., ChaCha, AES-based) reach GB/s and are “fast enough” even for games.
  • Detailed benchmarks (Go, multiple CPUs) show ChaCha variants within a small constant factor of PCG/xoshiro for typical use, especially with internal buffering.
  • Counter‑arguments emphasize latency vs throughput, instruction‑level parallelism, determinism with many RNG instances, and the large state needed to run CSPRNGs efficiently.
  • No clear consensus: CSPRNGs may now be viable for many games, but non‑crypto PRNGs remain simpler, smaller, and still significantly faster in some regimes.

Game Exploits and Anarchy Servers

  • Many praise the exploit as “art” and one of the most significant in anarchy Minecraft, though financially minor compared to real‑world RNG failures.
  • Discussion of 2b2t and similar servers: heavy metagaming, logistics, spying, and bug exploitation; some liken it to real warfare, others find it tedious.
  • PvP meta described as dominated by duped items, ender crystal spam, and automated hacked clients rather than vanilla combat mechanics.
  • Social downsides noted: pervasive slurs in chat; suggested mitigations include client‑side filters or moving coordination to external platforms.

Other RNG-Cracking Anecdotes

  • Historical examples: rock‑paper‑scissors bots that reverse PRNG state, fork‑based bots that try all moves, and code that rewrites stack history to “fix” past outcomes.
  • An online poker site once documented its PRNG for transparency; someone exploited this, ultimately helping improve their security.
  • Proposed fairness scheme: commit publicly to a daily seed (e.g., via hash), use it in a CSPRNG, and reveal later so players can verify non‑cheating.

LLL, Cryptography, and In‑Game RNG Manipulation

  • LLL lattice reduction is linked to other recent attacks (e.g., biased nonces in PuTTY, truncated LCG recovery, ECDSA key leaks).
  • Multiple comments praise the clarity of the LLL explanation and point to step‑by‑step treatments on Q&A sites.
  • Related Minecraft feats: farms and contraptions that deterministically force “random” lightning or maximum block drops, and even in‑game machinery predicting RNG state from TNT behavior.
  • Some participants frame all of this as “black magic” but also as highly inspirational for learning reverse engineering and crypto.