GPU-Accelerated Rule Processor & Generator

Concentrator v3.0

Generate, extract, and validate high-probability Hashcat rules using OpenCL, Markov models, and combinatorial mathematics.

OpenCL GPU Markov Models Interactive Mode Combinatorial Math BFS Chain Discovery

Overview

Concentrator v3.0 combines GPU-accelerated rule processing with statistical modeling to discover and validate password transformation rules from real-world corpora.

Unlike static rule lists, Concentrator derives rules from data — using Markov chain analysis to identify high-probability character transformations, then validates candidates via GPU to ensure they actually crack more hashes.

Core Features

⚡

OpenCL GPU acceleration

Massively parallel rule processing and candidate validation using the OpenCL kernel.

Markov chain modeling

Statistical n-gram analysis identifies the highest-probability character transformations.

⤳

Interactive mode

Live session for iterative exploration — tweak parameters and see results in real time.

✓

Config confirmation UX

Shows full config summary before starting — no surprise long runs.

Processing Pipeline

Step 1

Load corpus & rules

Memory-mapped loading of wordlist and existing rule files.

Step 2

Markov analysis

Build n-gram transition model of character frequency patterns.

Step 3

Candidate synthesis

Generate rule candidates from high-probability transitions and BFS chains.

Step 4

GPU validation

OpenCL kernel tests each candidate against hash corpus in parallel.

Step 5

Score & rank

Filter by hit rate threshold, rank by coverage.

Step 6

Export

Download ranked ruleset ready for Hashcat.

Operating Modes

⚡

Batch mode

Fully automated pipeline from corpus to ranked ruleset.

⤳

Interactive mode

REPL loop — adjust depth, threshold, and chain length live.

≡

Extraction mode

Derive rules directly from known plaintext/hash pairs.

Technical Details

Concentrator uses the shared OpenCL kernel that underpins Ranker and Aether — ensuring consistent rule semantics across the whole suite.

The Markov model supports n-gram orders 1–5. Higher orders capture more context but require more corpus data for reliable statistics.