Shows that the topology of causal knowledge graphs built from cipher behaviour carries enough structure to automate security-margin analysis and to identify ciphers blindly — without access to keys, plaintexts, or implementation details. Presented at ICECET 2026 in Rome.