We present IDRE (Integer-Dependent Receiver Encoding), a cryptographic protocol designed for sovereign, pre-provisioned networks where third-party trust infrastructure (PKI, certificate authorities) is unacceptable. Security emerges from the non-exportable geometry of internal vector fields rather than from stored or negotiated keys. Unlike traditional PKI-based protocols, IDRE transmits only semantic-free integers that act as pointers to a shared, dynamic internal state; intercepted traffic is statistically indistinguishable from random noise. We formalize the decoding function Mₜ = (Sₜ, Iₜ) (Sₜ) and prove that confidentiality holds under the Random Oracle Model with min-entropy assumptions on the field configuration. The protocol architecture draws on the functional connectivity of biological neural networks: each node maintains a private geometric attractor field, and communication succeeds only when two fields resonate at identical configurations analogous to synaptic coherence between cortical regions. IDRE was developed primarily as the security layer for MTI-HIVE, a platform for sovereign AI and local-first shared consensus between autonomous nodes. Through rigorous isolation of the cryptographic layer from the underlying transport, the protocol ensures high-integrity knowledge transfer within decentralized clusters without reliance on external trust anchors. This paper describes the complete v2. 4 protocol architecture, including: (1) a field fingerprint scan that derives keying material from deterministic vector geometry, (2) a mandatory HMAC-SHA256 pepper trapdoor that defeats gradient-descent weight recovery (Appendix A. 3), (3) an authenticated wire format with freshness enforcement, (4) a one-time challenge handshake that prevents session fixation, (5) Epoch Anchors for rolling chain-hash integrity within sessions, (6) a session-bound Neural Codec for adaptive payload compression, (7) Ghost Topology (vector space folding) for geometric obfuscation, (8) a network-sovereign tokenizer (VocabCodec) that cryptographically binds each deployment to its chosen vocabulary, (9) offline sealed letters for sessionless communication via audio carrier (IDRE-Silence), and (10) Dark Mode single-packet authorization for network invisibility. We validate the protocol through 57 unit tests and an adversarial test suite comprising replay attacks, rate-limit DoS, gradient-descent cryptanalysis, packet-loss resilience, codec poisoning, consensus-fork divergence, topology folding verification, tamper detection, and timing analysis. We provide empirical attack logs demonstrating that without pepper, gradient optimization recovers 97. 9% of fingerprint bits in under 3 minutes; with pepper active, the same attack plateaus at 59. 7% — confirming the HMAC trapdoor as a mandatory defense. Results confirm: zero semantic correlation for unauthorized receivers, immediate lockout upon state divergence, and deterministic security properties that hold under all tested conditions. A live reference implementation is publicly accessible at https: //idre. mti-evo. online for security researchers to interact with the protocol, inspect wire format, and test attack vectors. The complete source code, test suite, and attack harness are available at https: //github. com/Gastroam/idreₚrotocol. Scope note: IDRE is not a replacement for TLS, Signal, or WireGuard on the public internet. It addresses a distinct threat model where nodes are physically provisioned and no online key exchange is desired.
Miguel Alejandro Morelo Bustamante (Sun,) studied this question.