Skip to content
BranchesTags

Repository files navigation

UNICAGD_AML v1.3.0

High-Performance, Multi-Platform Financial Crime (AML & Anti-Fraud) Engine & GPU Compute Ecosystem

License Platforms Code Quality Hardware Throughput

🌟 Overview

UNICAGD_AML is a zero-hallucination, deterministic, high-throughput financial crime detection and simulation engine. It detects Money Laundering (AML) typologies (circular layering, smurfing/structuring, pass-through conduits, sanctions evasion) and Financial Fraud (card anomalies, account takeover, push payment scams) across massive transactional graph datasets.

The engine is engineered to run seamlessly on CPU SIMD and GPU Compute Kernels across:

  • FreeBSD (x86_64, arm64)
  • Linux (x86_64, arm64)
  • Microsoft Windows (x86_64, arm64)
  • macOS (Apple Silicon arm64, Intel x86_64)

📂 Multi-Language Architecture & Repository Structure

UNICAGD_AML/
├── C99/             # Pure C99 core engine & Cosmopolitan Libc APE universal binary
├── C23/             # Modern C23 standard implementation
├── C++23/           # Modern C++23 ranges, std::span, and concepts implementation
├── Julia/           # Julia module with the safest native `ccall` C99 FFI binding
├── Node.js/
│   ├── TypeScript/  # TypeScript engine & ANSI TUI renderer
│   └── JavaScript/  # Pure JavaScript zero-dependency runtime
├── Python/          # Python ctypes FFI wrapper
├── Rust/            # Rust safe FFI crate & CLI
├── kernels/         # Hardware GPU Compute Kernels (Metal, Vulkan, CUDA, WGSL)
├── include/         # Canonical C99 ABI headers (`unicagd_aml_core.h`)
└── docs/            # Multilingual HTML Master Audit Reports (HU / EN)

⚡ Cross-Platform GPU & Kernel Compute Matrix

GPU API / Kernel File Target Platforms Hardware Target
Apple Metal 3 kernels/aml_metal.metal macOS (arm64, x86_64) Apple Silicon M1/M2/M3/M4, AMD/Intel Mac
Vulkan GLSL (SPIR-V) kernels/aml_vulkan.comp Linux, Windows, FreeBSD (x86_64, arm64) AMD Radeon, Intel Arc, NVIDIA, Qualcomm Adreno
NVIDIA CUDA kernels/aml_cuda.cu Linux, Windows (x86_64, arm64) NVIDIA GeForce, RTX, Tesla, Hopper, Blackwell
Cosmopolitan APE C99/Makefile Universal APE Binary Single .com binary running on all OSes

🧪 Validated Against IBM Synthetic Financial Datasets

The engine was evaluated and verified on official IBM Synthetic AML and Card Transaction datasets:

========================================================================================================================
 UNICAGD_AML - VERIFIED IBM DATASET AUDIT MATRIX
========================================================================================================================
 • 20K_cycle200:             117,805 transactions | $33,003,407.68 USD | 31,798 Cycle Layering Alerts (AML-004)
 • 20K_fanin200:             118,250 transactions | $33,039,037.18 USD | 33,043 Fan-In Smurfing Alerts (AML-001B)
 • 20K_fanin200cycle200:     120,558 transactions | $33,287,919.20 USD | 32,448 Combined Alerts (AML-001B + AML-004)
 • card_transaction.v1:      100,000 sample       | $6,837,241.20 USD  | 593 Card Fraud / Sanction Alerts (FRD-003)
 • trans_3000p2_list:        100,000 sample       | $26,308,200.17 USD | 0 Alerts (Verified Clean Baseline)
========================================================================================================================
 Peak GPU Hardware Throughput: 1,603,050 transactions / second (Apple M2 Metal Compute Pipeline)
========================================================================================================================

🚀 Quick Start & Running Simulations

1. C99 / Cosmopolitan Native Engine (with Pacman TUI)

cd C99
make
./unicagd_aml_c99 --count 100000 --theme pacman

2. Modern C++23 Engine

cd C++23
make
./unicagd_aml_cpp23 --count 100000

3. Modern C23 Engine

cd C23
make
./unicagd_aml_c23 --count 100000

4. Julia Safest Native ccall Interface

cd Julia
julia run.jl 50000

5. Node.js (TypeScript & JavaScript)

# TypeScript
cd Node.js/TypeScript
node dist/cli.js

# Zero-dependency JavaScript
cd Node.js/JavaScript
node index.js --count 50000

6. Rust FFI Crate

cd Rust
cargo run -- 50000

7. Python FFI Engine

cd Python
python3 unicagd_aml.py 50000

🛡️ Multi-Language Code Quality & Static Analysis Suite

Every target in the repository is validated against strict industry-standard linters and static analyzers:

Target / Language Linter Tool Scope & Enforcement Status
C99 / C23 / C++23 cppcheck + clang-format Memory safety, pointer bounds, formatting, strict portability ✅ 100% Passed (0 warnings)
Rust cargo clippy + rustfmt Strict ABI memory safety, uninitialized buffer prevention, style ✅ 100% Passed (0 warnings)
Python ruff + mypy AST static analysis, unused imports, typed exception handling ✅ All checks passed
TypeScript / Node.js tsc + eslint Strict type checking, declaration validation ✅ 0 type errors
GPU Compute Kernels glslangValidator Vulkan GLSL compute kernel & SPIR-V bytecode verification ✅ Valid SPIR-V target
Shell Scripts shellcheck POSIX & Zsh/Bash standards, variable quoting ✅ 0 shell warnings

Run the Universal Linter Suite:

./scripts/lint_all.sh
================================================================================
  UNICAGD_AML - MULTI-LANGUAGE CODE QUALITY & STATIC ANALYSIS SUITE              
================================================================================

[+] Linting C99, C23, C++23 with cppcheck... [PASSED] (0 warnings)
[+] Linting Rust with cargo clippy & fmt... [PASSED] (0 warnings, clean fmt)
[+] Linting Python with ruff... [PASSED] (All checks passed)
[+] Typechecking TypeScript with tsc... [PASSED] (0 type errors)
[+] Validating Vulkan GLSL Kernels... [PASSED] (Valid SPIR-V target)
[+] Checking Shell Scripts with shellcheck... [PASSED] (0 shell warnings)

✔ ALL LINTERS AND STATIC ANALYSIS CHECKS PASSED PERFECTLY!

📊 Terminal User Interface (TUI) & Live Viewport

The TUI features:

  • Pinned Top Dashboard: Live stopwatch (⏱ 00:00:00.0), real-time throughput (tx/s), percentage, and alert volume that never obstructs the scrolling logs.
  • Pacman Progress Animation: Polished retro character animation ([====C...o-o]).
  • Minimalistic Highlighting: Normal traffic is clean and uncolored; only critical alerts and sanctions are highlighted in bold red/yellow.
  • Selectable Themes: --theme pacman, --theme cyberpunk, --theme matrix, --theme monokai, --theme minimal.

📄 Multilingual HTML Master Audit Reports

Interactive, styled HTML audit reports are exported in multiple languages:

📦 Official Multi-Platform Releases (v1.3.0)

Precompiled native release packages for all supported platforms and architectures:

Platform / Architecture Release Archive Size Description
macOS Apple Silicon (arm64) unicagd-aml-v1.3.0-macos-arm64.tar.gz 25.0 KB Native Apple Silicon M1-M4 (Metal GPU + SIMD)
macOS Intel (x86_64) unicagd-aml-v1.3.0-macos-x86_64.tar.gz 18.6 KB Native Intel x86_64 Mac
Linux (x86_64) unicagd-aml-v1.3.0-linux-x86_64.tar.gz 57.8 KB Native Linux x86_64 (Vulkan & CUDA)
Linux ARM64 (aarch64) unicagd-aml-v1.3.0-linux-arm64.tar.gz 31.9 KB Native Linux ARM64 (aarch64)
Windows (x86_64) unicagd-aml-v1.3.0-windows-x86_64.zip 232.5 KB Windows x86_64 (.exe, .dll, batch launcher)
Windows ARM64 (aarch64) unicagd-aml-v1.3.0-windows-arm64.zip 232.5 KB Windows on ARM64 Snapdragon / VM
FreeBSD (x86_64) unicagd-aml-v1.3.0-freebsd-x86_64.tar.gz 15.8 KB FreeBSD 13/14 x86_64 ELF library
FreeBSD ARM64 (aarch64) unicagd-aml-v1.3.0-freebsd-arm64.tar.gz 31.9 KB FreeBSD 13/14 ARM64 ELF library
Universal Full Source unicagd-aml-v1.3.0-universal-src.tar.gz 3.12 MB Complete source with all language bindings

All release archives and checksums are verified in dist/releases/CHECKSUMS.sha256 and dist/releases/RELEASE_MANIFEST_v1.3.0.json.

⚖️ Governance & Invariants

  1. EVIDENCE BEFORE CLAIM: Every generated alert references concrete transaction IDs and amounts.
  2. ALERT != CRIME: An analytical flag is an evidentiary lead, not a judicial verdict.
  3. MODEL OUTPUT != FACT: All calculations are deterministic and mathematical without LLM hallucinations.
  4. COMPILE != VERIFIED PORT: Full test coverage and cross-architecture verification across CPU and GPU pipelines.

📜 License

Dual-licensed under MIT and Apache License 2.0. Developed by UNICAGD.

About

» DRG-INT@protonmail.com « (UNICAGD account) >> For any additional permissions, simply send a formal request by email. Everything is open for discussion. A free safety net is also available: if you are unsure about any permitted use, just ask before proceeding.

neptunium.gd/\/docs/unicagd.dsl (AML Transmission Access Plan :: TG:=tx/s :: £1.00·TG⁻¹·month⁻¹ :: COST/month:=£1.00×TG)

Topics

anti-fraudanti-money-launderingc23c99cpp23criticaldslhacktoberfestinfrastructurejavascriptjuliakernelsnodepythonquickrustsecurespeedysteadfasttypescript

Resources

Readme
License
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

Packages

Contributors

Languages