convez.ai
MCP ServerSegurançaAvançado
mdaminxf

DigiSec - Plataforma de Perícia Digital Autônoma

mdaminxf/DigiSec

Forense digital autônoma com chain-of-custody garantida e relatórios judiciais

Ver no GitHub
InicialPluginScore · de 0 a 100

MCP server especializadicado em forense digital (DFIR) que encapsula ferramentas SIFT com garantia de zero-violação de evidências. Integra Volatility 3, análise de artefatos de disco ($MFT, Prefetch, Evtx) e correlação multi-fonte em timelines unificadas, gerando relatórios em PDF prontos para tribunal via extensões YAML sem código.

Instalação

$npx convez-ai mdaminxf/DigiSec

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Investigação de incidentes com segurança de evidências
  • Análise de memória e artefatos de disco correlacionados
  • Geração de relatórios forenses prontos para audiência

Veredito

Solução especializada e bem arquitetada para DFIR que resolve problemas reais de chain-of-custody e hallucination em contextos forenses. Exige expertise técnica significativa, mas oferece garantias críticas que ferramentas genéricas não têm.

README

A tradução para o português deste README ainda está na fila.

Extensible DFIR MCP Platform (DigiSec)

logo

SANS Hackathon Submission

An autonomous, zero-spoliation digital forensics platform. Wrap SIFT's tools into a type-safe MCP server, execute multi-source correlation, dynamically load zero-code YAML extensions, and output courtroom-ready PDF reports.


🏆 Hackathon Ideas Implemented

This architecture natively solves several of the hardest problems in AI-assisted DFIR:

  1. The Purpose-Built MCP Server (Zero Spoliation): LLMs do not get raw bash access. They must route through our FastMCP server. Disk evidence is safely exposed via mount_e01_image which enforces OS-level ntfs-3g ro,loop read-only mounts. Evidence cannot be modified.
  2. The Self-Correcting Triage Agent (Anti-Hallucination): The Detection Engine heuristics strictly enforce the DFIR standard of Observed Fact → Evidence → Inference → Conclusion. Findings natively return requires_validation=True and alternative_explanations to prevent the AI from jumping to speculative conclusions.
  3. Multi-Source Correlation Engine: Seamlessly merges memory extraction (Volatility 3) and disk artifacts ($MFT, Prefetch, Evtx) into unified timelines.
  4. Zero-Code Extensibility: The Dynamic Plugin Loader instantly compiles any .yaml or .py files dropped into the plugins/ directory into type-safe AI tools without requiring server restarts.

🏗 Architecture

The platform acts as an unbreachable wall between the AI and the raw evidence, guaranteeing chain-of-custody integrity while granting the LLM massive analytical power.

graph LR

%% =========================
%% STYLES
%% =========================
classDef llm fill:#4a148c,stroke:#fff,stroke-width:2px,color:#fff
classDef tools fill:#0d47a1,stroke:#fff,stroke-width:1px,color:#fff
classDef safe fill:#1b5e20,stroke:#fff,stroke-width:1px,color:#fff
classDef report fill:#b71c1c,stroke:#fff,stroke-width:2px,color:#fff
classDef isolate fill:#f57f17,stroke:#fff,stroke-width:2px,color:#fff

%% =========================
%% ROOT
%% =========================
A([Prompt])-->B[Orchestrator]:::llm

%% =========================
%% TOOLS FRAME
%% =========================
subgraph T[TOOLS]
direction TB
C1[Isolation]:::isolate
C2[Intel]:::tools
C3[Disk]:::tools
C4[Memory]:::tools
end

B-->C1
B-->C2
B-->C3
B-->C4

%% =========================
%% ISOLATION FRAME
%% =========================
subgraph I[ISOLATION]
direction TB
D1[Scope]
D2[Case]
end

C1-->D1
C1-->D2

%% =========================
%% DISK FRAME
%% =========================
subgraph D[DISK]
direction TB
D3[Mount]-->D4[MFT]-->D5[Norm]-->E1[DiskOut]:::safe
end

C3-->D3

%% =========================
%% MEMORY FRAME
%% =========================
subgraph M[MEM]
direction TB
D6[Snap]-->D7[Proc]-->D8[Net]-->E2[MemOut]:::safe
end

C4-->D6

%% =========================
%% INTEL FRAME
%% =========================
subgraph TI[INTEL]
direction TB
D9[IOC]-->D10[Rep]-->D11[MITRE]-->E3[IntelOut]:::safe
end

C2-->D9

%% =========================
%% CORRELATION FRAME
%% =========================
subgraph C[CORR]
direction TB
E1-->G[Merge]
E2-->G
E3-->G
G-->G1[Anom]-->G2[Valid]-->G3[Find]
end

%% =========================
%% REPORT FRAME
%% =========================
subgraph R[REPORT]
direction TB
H1[Struct]-->H2[Sum]-->H3[Render]-->H4([Final]):::report
end

G3-->H1

🚀 Setup & Execution

1. Requirements

  • Python 3.14+
  • ewf-tools and ntfs-3g (for .e01 disk mounting)
  • weasyprint (for PDF generation)

2. Run the Server

The environment is pre-configured. Start the FastMCP server:

source venv/bin/activate
python server.py

3. Claude Desktop Configuration

Add the server to your claude_desktop_config.json:

{
  "mcpServers": {
    "digisec-mcp-server": {
      "command": "/home/sansforensics/Desktop/sift-mcp/venv/bin/python",
      "args": [
        "/home/sansforensics/Desktop/sift-mcp/server.py"
      ]
    }
  }
}

🧩 Adding New Tools (Zero-Code YAML)

You do not need to know Python to extend this platform. Simply drop a .yaml file into the plugins/ directory.

Example plugins/network/check_ip.yaml:

name: check_ip_reputation
description: "Checks the reputation of an IP address using curl."
arguments:
  ip_address:
    type: str
command: "curl -s 'https://api.threatintel.com/ip/{ip_address}'"

The server will dynamically generate a Python tool, inject the arguments safely, and register it with the AI automatically.

📄 Executive PDF Reporting

To prevent massive walls of unreadable text in the chat interface, the platform includes a WeasyPrint renderer (plugins/analysis/custom_report.py). The LLM will automatically generate elegant, styled HTML and compile it into a confidential PDF document saved directly to your workspace.

DigiSec

Ver o original no GitHub

Bom para

Relacionados