KVKK Compliance: Anonymization + Right to Erasure + Machine Unlearning
KVKK Article 7: 'Right to Erasure'. Citizen says 'delete me from dataset': re-train expensive (millions). **Machine Unlearning** alternative: SISA approach or gradient ascent method. KVKK Board decisions, practical example (TR banking citizen erasure request).
Şükrü Yusuf KAYA
30 min read
Advanced
1. Machine Unlearning Yöntemleri#
| Method | Cost | Etki |
|---|---|---|
| Full retrain | $$$$ (her vatandaş silme talebinde tam retrain) | Tam |
| SISA (Bourtoule 2021) | $$ | Shard re-train (sadece etkilenen shard) |
| Gradient ascent | $ | Hızlı, ama imperfect |
| Knowledge distillation | $$ | Yeniden train ama küçük dataset |
SISA mantığı: Dataset'i N shard'a böl, her shard ayrı model eğit, inference'te ensemble. Bir kişi sil → sadece o kişinin olduğu shard'ı retrain.
Cookbook'un kuralı: TR production'da KVKK silme hakkı için SISA pre-design yap. Sonradan unlearning impossible.
✅ Teslim
- KVKK Madde 7 oku. 2) SISA paper'ı incele. 3) Sonraki ders: 18.3 — Model Lisans Labirenti.
Yorumlar & Soru-Cevap
(0)Yorum yazmak için giriş yap.
Yorumlar yükleniyor...
Related Content
Part 0 — Engineering Foundations
Welcome to the Fine-Tuning Cookbook: System, Stage Taxonomy, and the Reproducibility Contract
Start LearningPart 0 — Engineering Foundations
Reproducibility Stack: Seeds, cuDNN Flags, and Deterministic CUDA — End the 'Works on My Machine' Problem
Start LearningPart 0 — Engineering Foundations