Hi everyone! As part of an ongoing research series exploring censorship-resistant peer-to-peer overlay architectures (following The Invariant Boundary Theorem and The ECDH Protocol Discipline Trap), I have formalized the next critical layer: Post-Quantum Stateless Steganography.
The preprint is officially available on Zenodo: 
DOI: 10.5281/zenodo.23437773
The Problem: Why Naive Post-Quantum Migration Fails for P2P Tunnels
Transitioning censorship-resistant overlays (such as those protecting decentralized validator communication or private RPC relays) from classical Elliptic Curves (ECDH / X25519) to post-quantum primitives (like NIST ML-KEM and ML-DSA) breaks existing steganographic tunnel designs under two mathematical and systems-level traps:
-
The Modulo Bias Invariant ($q \neq 2^k$): Standard lattice KEMs operate over prime moduli (e.g., $q = 3329$ in ML-KEM). Naive bit-packing leaves upper sub-intervals unpopulated, driving statistical distance from uniform noise ($\Delta(\mathcal{P}, \mathrm{U}) \to 1$). Under information-theoretic bounds, this automatically compiles into a zero-cost $\mathcal{O}(1)$ passive Deep Packet Inspection (DPI) line-rate distinguisher.
-
The MTU Fragmentation Trap: Composite post-quantum 0-RTT envelopes (KEM ciphertext + signature + AEAD) exceed $4.4$~KB, forcing multi-packet fragmentation. Because kernel-bypass filters like eBPF/XDP inspect individual frames at the driver level, fragmented 0-RTT payloads force the responder to allocate stateful IP/UDP reassembly buffers prior to AEAD verification—re-opening memory-exhaustion DoS vectors and multi-packet dimensional fingerprints.
Proposed Architectural Solutions
To preserve strict $\mathcal{O}(1)$ kernel-bypass Silent Drop ($\bot$) and Silent Session Transition guarantees without leaking stateful fingerprints, the paper establishes:
- Algebraic Uniformization: Leveraging native power-of-two Learning With Rounding (LWR) lattices (e.g., Saber/FrodoKEM) or constant-time Inverse Rejection Sampling to map prime-modulus ciphertexts into indistinguishable uniform bit strings ($\Delta(\mathcal{P}, \mathrm{U}) \le 2^{-\lambda}$).
- Sub-MTU 0-RTT Envelope Architectures: Designing compact post-quantum envelopes ($\le 1232$~bytes for IPv6 / $\le 1472$~bytes for IPv4) via:
- HNDL-Resilient Hybrid Envelopes (Post-quantum KEM + obfuscated classical signature for real-time authentication).
- Pure Post-Quantum Sub-MTU Envelopes (Utilizing uniformized NTRU signatures like Falcon-512 or unilateral Authenticated KEMs / Signcryption).
- Fujisaki–Okamoto Implicit Rejection: Exploiting IND-CCA2 KEM implicit rejection so that random DPI probes or cross-node replays silently evaluate to pseudorandom keys in constant time, failing AEAD verification and triggering instant silent drops.
Relevance to Ethereum & P2P Protocols
As Ethereum validator networks, libp2p-based gossipsub overlays, and RPC transport layers prepare for the post-quantum era against Harvest-Now-Decrypt-Later (HNDL) threats, maintaining low-overhead, un-fingerprintable wire envelopes at the network driver level is essential.
I welcome your thoughts, security reviews, and feedback on applying these sub-MTU lattice uniformization techniques to p2p transport security!