rednblue 3.0.0

Zero-Knowledge Adversarial Security Testing for AI Models

RednBlue CLI v3.0.0

Zero-Knowledge Adversarial Security Testing for AI Models

RednBlue CLI is a command-line tool for testing the adversarial robustness of machine learning models. Run security assessments locally — your model never leaves your infrastructure.

███████████               ███████████
▒▒███▒▒▒▒▒███             ▒▒███▒▒▒▒▒███
 ▒███    ▒███  ████████    ▒███    ▒███
 ▒██████████  ▒▒███ ▒▒███  ▒██████████
 ▒███▒▒▒▒▒███  ▒███  ▒███  ▒███▒▒▒▒▒███
 ▒███    ▒███  ▒███  ▒███  ▒███    ▒███
 █████   █████ ████  █████ ███████████
▒▒▒▒▒   ▒▒▒▒▒ ▒▒▒▒  ▒▒▒▒▒ ▒▒▒▒▒▒▒▒▒▒▒

Zero-Knowledge Adversarial Security Testing

---

[3.0.0] - 2026-0412

🚀 New Features

Desktop UI

• New rnb ui command - Launch a graphical interface for non-CLI users
• Built with Eel framework (Python-JavaScript bridge)
• Light theme matching the RednBlue Dashboard design
• Real-time progress tracking with per-dimension status
• Interactive radar chart visualization (Chart.js)
• Token validation with credit balance display
• Multi-language support (English, French)

Auto GPU Detection

• Automatically selects the best available compute device
• Priority: CUDA → MPS (Apple Silicon) → CPU
• No manual --device flag needed

Multi-Architecture Classifier Support

Expanded model architecture detection to support:

Architecture	Variants
ResNet	18, 34, 50, 101, 152
VGG	11, 13, 16, 19
DenseNet	121, 161, 169, 201
EfficientNet	b0, b1, b2, b3, b4, b5, b6, b7
MobileNet	v2, v3_small, v3_large
Inception	v3
SqueezeNet	1_0, 1_1
ShuffleNet	v2
GoogLeNet	-
AlexNet	-

🐛 Bug Fixes

• Fixed tier parameter not passed to classifier attacks - Enterprise tokens now correctly run all 8 attacks instead of defaulting to Freelancer's 5
• Fixed input_path undefined error - Added missing Path(input_dir) conversion in run_preview()
• Fixed Fore not defined error - Added missing colorama import in run_preview()
• Fixed token validation - Now correctly uses APIClient.validate_subscription() to fetch credit info

✨ Improvements

• Better error messages - Clear explanations when attempting to test unsupported model types (segmentation/detection models)
• CEO-friendly dimension labels - Display human-readable names instead of attack codes (e.g., "Noise Resilience" instead of "GNI")
• Cleaner console output - Tier and device info displayed at test start

📁 New Files

rnb/
├── ui.py              # Desktop UI launcher
└── web/
    ├── index.html     # UI layout (Tailwind CSS)
    ├── app.js         # UI logic & Chart.js radar
    ├── i18n.js        # Internationalization (EN/FR)
    └── logo.png       # RednBlue logo

⚠️ Known Limitations

Models not supported for testing (different output types):

Model Type	Examples	Reason
Segmentation	DeepLabV3, FCN	Output is mask, not class
Detection	Faster-RCNN, RetinaNet, SSD	Output is boxes, not class
Keras	.h5 files	TensorFlow format, not PyTorch

For YOLO detection models, select "YOLO" as model type.

---

Features

• Zero-Knowledge Protocol — Model weights and data never leave your infrastructure
• Image Classifiers — Test ResNet, VGG, EfficientNet, and custom architectures
• YOLO Detection — Full support for YOLOv5, YOLOv8, YOLOv10, YOLOv11
• Tier-Based Testing — Freelancer (quick scan) and Enterprise (comprehensive)
• Encrypted Submission — AES-256 encrypted results with HMAC-SHA256 signing
• Multi-Jurisdiction Compliance — EU AI Act, NIST AI RMF, ISO/IEC 42001, UK DSIT, Canada AIDA, Singapore MAIGF

---

Installation

# Install from PyPI
pip install rednblue

# Verify installation
rnb

---

Requirements

• Python 3.8+
• PyTorch 2.0+
• CUDA (optional, for GPU acceleration)

---

Quick Start

1. Set your token

# Windows
set RNB_TOKEN=RB-XXXXXX-YYYYYY

# Linux/Mac
export RNB_TOKEN=RB-XXXXXX-YYYYYY

2. Run a security assessment

Image Classifier:

rnb preview --model resnet50.pth --input ./test_images --model-type classifier

YOLO Detection Model:

rnb preview --model yolov10n.pt --input ./test_images --model-type yolo

3. Submit for certification

rnb preview --model yolov10n.pt --input ./images --model-type yolo --submit

---

Commands

Command	Description
rnb	Show welcome banner and quick start
rnb preview --help	Run adversarial attacks
rnb status	Check token validity and tier
rnb optimize-epsilon	Optimize epsilon values (Enterprise)
rnb test-llm	Test LLM models (Enterprise)

---

Assessment Dimensions

Classifier Models

Dimension	Description
Noise Resilience	Stability under sensor noise and interference
Spatial Consistency	Robustness to spatial feature shifts
Universal Pattern Defense	Resistance to universal perturbation patterns
Feature Stability	Internal representation integrity
Confidence Calibration	Prediction reliability accuracy
Iterative Stress Tolerance	Defense against sustained pressure
Optimization Attack Defense	Resistance to optimized adversarial inputs
Deep Perturbation Resistance	Resilience against deep layer perturbations

YOLO Detection Models

Dimension	Description
Noise Resilience	Stability under sensor noise
Input Perturbation Defense	Resistance to subtle input modifications
Iterative Stress Tolerance	Defense against multi-step attacks
Detection Consistency	Reliable detection under varying conditions
Targeted Evasion Defense	Resistance to deliberate misclassification
Object Persistence	Maintains detections under perturbations
Multi-Object Stability	Accuracy in crowded scenes
Black-Box Resilience	Defense without model access
Query-Limited Defense	Resistance to low-query probing

---

Tier Comparison

Feature	Freelancer	Enterprise
Classifier Attacks	5	8
YOLO Attacks	4	9
Epsilon Values	2	4
Total Scenarios	~10-20	~30-70
LLM Testing	❌	✅
Epsilon Optimization	❌	✅

---

Output Example

============================================================
  RednBlue Security Preview — YOLO Detection
============================================================
  Attacks run    : 21
  Successful hits: 0/21 (0%)
  Robustness rate: 100%
  Estimated Grade: GOLD

  ⚠️  This is a preview only
  → Visit: https://rednblue.io/checkout
  → Re-run with: rnb preview --model-type yolo --submit

---

Certification Grades

Grade	Score	Meaning
🥇 GOLD	≥90%	Excellent robustness, deployment ready
🥈 SILVER	≥75%	Good robustness, minor improvements recommended
🥉 BRONZE	≥50%	Moderate robustness, improvements needed

---

Architecture

---

Links

• Platform: https://dashboard.rednblue.io
• Documentation: https://docs.rednblue.ai
• Website: https://rednblue.io

---

Authors

• Dr. Mahdi Deramgozin — Chief AI Officer
• Dr. Saeid Samizade — Chief Technology Officer

---

License

Proprietary — RednBlue SAS © 2026

Made in France 🇫🇷