k-anonymity protects against one kind of disclosure but not another — what's the blind spot?

k-anonymity protects identity (you can't tell which person), but NOT attribute privacy (you may still learn their sensitive value).

This is the single most important limitation. k-anonymity guarantees an attacker can't pin a record to one person — it defends against identity disclosure. But it says nothing about the sensitive attribute inside an equivalence class. If the group's sensitive values aren't varied, the attacker can learn the secret without ever identifying the individual — that's attribute disclosure.

This blind spot is exactly what the homogeneity attack exploits, and what l-diversity was created to close.

Tip: "Who is this?" vs "What's true about them?" — k-anonymity only answers the first. Privacy needs both protected.