Sharing some practical experience evaluating G-CTR-like guarantees from a security perspective.
When adversaries adapt, several assumptions behind the guarantees degrade faster than expected. In particular:
- threat models get implicitly frozen
- test-time confidence doesn’t transfer to live systems
- some failures are invisible until exploited
Curious if others in netsec have seen similar gaps between formal assurance and operational reality.