Feat/mea attack6

[razon1494]

📦 Pull Request Template

Thank you for your contribution! Please complete the checklist and provide relevant details below to help us review your PR effectively.

---

📋 Summary

What

• Add ModelExtractionAttack6 in models/attack/mea/MEA.py.
• Re-export it via models/attack/__init__.py so it’s importable as from models.attack import ModelExtractionAttack6.
• Make attack() return a results dict: {"fidelity": ..., "accuracy": ...} (per IMPLEMENTATION.md).
• Add runnable examples:
  • examples/example.py — runs Attack-6 on Cora.
  • examples/mea_cora_attack6.py — helper script used while testing.
  • examples/defense_cora.py — defense quick-start used during validation.
• Housekeeping: ignore *.egg-info build artifacts.

Why

• Integrates my reproduction into the modular BaseAttack pattern, using the project Dataset and device handling with minimal changes.

How to run

# inside venv, after installing deps per README (PowerShell)
$env:DGLBACKEND="pytorch"
$env:PYGIP_DEVICE="cpu"
python examples/example.py

---

🧪 Related Issues

• N/A (happy to link if there’s a tracking issue)

---

✅ Checklist

• My code follows the project's coding style
• I have tested the changes and verified that they work
• I have added necessary documentation (if applicable)
• I have linked related issues above (if any)
• The PR is made from a feature branch, not main (feat/mea-attack6)

---

🧠 Additional Context (Optional)

Environment

• Windows 64-bit, Python 3.11.9
• PyTorch 2.3.0+cpu, DGL 2.2.1, NumPy 1.26.4

Notes

• DGL prints a deprecation warning about DGLGraph(...); left as-is for minimal change.
  I can switch to dgl.from_networkx(...) + dgl.add_self_loop(...) in a follow-up if preferred.
• If updating reqs.txt isn’t desired, I can revert that line.