🎯 MOSAIC — v1.0.0-alpha

Behavioral Intelligence Across Platforms

A privacy-first, multi-platform framework for behavioral and sociodynamic analysis

📺 Demo

🧭 What is MOSAIC?

MOSAIC is not a simple OSINT aggregator. It is a behavioral analysis framework designed to help researchers, analysts, and security professionals observe, structure, and interpret behavioral signals across multiple online platforms.

MOSAIC collects heterogeneous public data and organizes it into behavioral dimensions, enabling comparative and longitudinal analysis that is not possible with single-source approaches.

⚠️ MOSAIC does not produce psychological diagnoses or absolute truths. It provides structured signals and analytical hypotheses that must be interpreted by a human analyst.

🎯 Vision

Most tools focus on what a person publishes. MOSAIC focuses on how behavior emerges across environments.

Instead of aggregating profiles, MOSAIC builds behavioral fingerprints across three complementary dimensions:

• Technical Skills, expertise, problem-solving strategies (GitHub, StackOverflow)

• Social Interactions, affiliations, discourse patterns, community dynamics (Telegram, Reddit, Mastodon, Bluesky)

• Influence Content production, themes, visibility, narrative consistency (YouTube, Medium)

This multi-angle approach enables sociodynamic, security, and credibility analyses that remain contextual, explainable, and privacy-aware.

🚀 Quick Start

Installation

git clone https://github.com/Or1un/MOSAIC.git
cd MOSAIC
pip install -r requirements.txt

Configuration (Optional)

# Most platforms work without credentials
# Some sources (YouTube, Telegram) require API keys
cp modules/config.yaml.example modules/config.yaml
nano modules/config.yaml

Run

python3 mosaic.py

🤖 LLM Setup

1. Install Python dependencies

pip install -r requirements_llm.txt

2. Install Ollama server (one-time)

# Linux/macOS
curl -fsSL https://ollama.com/install.sh | sh

# Windows
# Download from https://ollama.com/download

3. Start Ollama & pull models - Qwen:0.5b (PoC)

ollama serve &
ollama pull qwen:0.5b

🌍 Use Cases

🎯 Talent & Expertise Assessment

Go beyond resumes by analyzing:

• technical depth
• communication style
• peer recognition
• community engagement

Example: Assess a senior DevSecOps profile via GitHub activity (technical rigor), StackOverflow answers (problem-solving style), and Medium publications (knowledge transmission).

🔒 Digital Footprint & OPSEC Analysis

Identify exposure risks, oversharing patterns, and inconsistencies in public presence.

Example: Audit the online footprint of an executive or security-sensitive role to detect potential attack surfaces or reputational risks.

🕵️ Due Diligence & Credibility Analysis

Cross-validate claims and behavioral consistency across platforms.

Example: Before a partnership or investment, evaluate whether public behavior aligns with stated expertise and responsibilities.

🔬 Behavioral & Sociodynamic Research

Study:

• discourse evolution
• expertise diffusion
• community influence patterns
• cross-platform behavioral shifts

Example: Analyze how technical authority translates (or not) into social influence across different ecosystems.

🧩 Core Workflow

1️⃣ Collect — Multi-Source Signals

MOSAIC gathers heterogeneous public data to avoid mono-platform bias.

Currently supported sources:

• Technical: GitHub, StackOverflow
• Social: Telegram, Reddit, Mastodon, Bluesky
• Influence: YouTube, Medium

💡 MOSAIC works best when you already know where your target is active. Focused collection improves signal quality and reduces noise.

Future sources (research ideas): professional platforms, forums, niche communities. Contributions and discussions are welcome.

2️⃣ Context — Analytical Framing

Analysis is guided by contextual prompts, not generic instructions.

Available contexts (Alpha):

• Recruitment.md — technical & soft skills assessment
• OPSEC_audit.md — digital footprint & exposure analysis
• Threat_Intelligence.md — attribution & consistency research
• PoC.md — quick exploratory analysis

Prompts are Markdown templates stored in modules/prompts/.

⚠️ Alpha note Current prompts prioritize architecture and extensibility over refinement. MOSAIC assumes analysts will adapt prompts to their methodology and domain.

3️⃣ Analysis — AI as an Assistant

MOSAIC supports two analysis modes:

✅ Local (Privacy-First — Recommended)

• Uses local LLMs via Ollama
• Data never leaves your machine
• Best for security-sensitive contexts

ollama pull mistral:7b-instruct

Supported models: mistral, llama, qwen, etc.

⚠️ Cloud (Optional)

• Higher performance
• User-managed privacy trade-offs
• Upload only exported JSON files from results/

💡 Best results are obtained when combining MOSAIC outputs with external context (CVs, interviews, references, timelines).

4️⃣ Insights — Human-Centered Interpretation

Insight quality depends on:

• data diversity
• prompt precision
• analyst expertise
• model capabilities

MOSAIC provides structure and signals. You provide judgment.

⚠️ Ethical & Legal Framework

MOSAIC is designed for legitimate, consensual, and lawful use.

❌ Not Intended For

• Surveillance or stalking
• Harassment or intimidation
• Unauthorized background checks
• Mass automated profiling
• Violations of platform Terms of Service

🔐 Privacy & Responsibility

• No mandatory cloud usage
• Local analysis by default
• Results stored locally (results/, .git-ignored)
• Users are responsible for GDPR and local compliance
• Credentials stored securely in config.yaml

By using MOSAIC, you agree to use it ethically, responsibly, and legally.

🤝 Community & Contributions

Questions, ideas, or research directions?

• 📧 Email: [email protected]
• 🐛 Issues: GitHub Issues
• 💡 Discussions: GitHub Discussions
• 📖 Wiki: coming soon

MOSAIC is a methodological framework — contributions that improve prompts, analysis logic, or ethical safeguards are especially welcome.

Made with 🔍 by Or1un