What are the "two core techniques" for protecting data, and which model represents each?

Generalizing/suppressing the data (k-anonymity family) and adding mathematical noise (differential privacy) — the two fundamental levers of anonymization.

At the highest level, data protection comes down to two moves:

• Generalizing the data — removing or replacing portions of data elements that could identify individuals (e.g. dropping area codes, broadening values). This is the world of k-anonymity, l-diversity, t-closeness.
• Adding noise to data — introducing mathematical noise that makes it hard to tell whether any individual is in the dataset. This is the world of differential privacy.

The first preserves real (but coarsened) values; the second perturbs values to provide deniability. Strong systems often combine them.

Tip: Blur it (generalize) or shake it (add noise) — almost every anonymization technique is a flavor of one of these two ideas.