The Phalanx: Authenticated Object Custody and Ciphertext-Feedback Research

Encrypted data received over an unreliable link presents two distinct questions: whether an individual record is authentic, and whether a complete object has been received. The Phalanx specifies a bounded object-custody protocol that answers these questions separately. Independently authenticated AES-256-GCM-SIV records can arrive in arbitrary order; an encrypted terminal manifest binds their ordered ciphertext transcript, count, geometry and total length. Plaintext is released only after complete-object verification. We give a conditional completion-integrity argument, an exact support bound for a separate ciphertext-feedback reference, and an analytical model for the cost of authenticated selective retransmission. The executable study includes 1,600 synthetic channel trials and 23,042 rejected malformed Seal inputs. At a bit error probability of 5 × 10−5 and a 512-byte payload chunk, selective receipt completed 40 of 40 trials, compared with 31 of 40 for a discard-and-restart policy using the same encrypted records and retry ceiling. The corresponding mean transmitted volumes were 6,706 and 34,243 bytes. The infinite-retry data-record model predicts a continuous optimum near 491 bytes for the measured 120-byte record overhead. All measurements, negative results, encodings and reproduction scripts are included. The contribution is a protocol composition and its reproducible analysis; it is not a new block cipher, an independent security certification or a claim of flight readiness.