Post-Quantum Stateless Steganography & MTU Bounds: Bridging P2P Censorship Resistance to Quantum Networks

What it is

A forum announcement by the author of a series on censorship-resistant P2P overlays (earlier parts: "The Invariant Boundary Theorem" and "The ECDH Protocol Discipline Trap"). It points to a Zenodo preprint, DOI 10.5281/zenodo.23437773. The post is a summary only. It gives no benchmarks, code, or measured distinguisher results, and the full argument is in the preprint, which I have not seen.

Details the catalog summary omits

• Concrete size budgets. Envelopes should fit in ≤1232 bytes for IPv6 and ≤1472 bytes for IPv4. That is the sub-MTU target for avoiding fragmentation. The composite envelope (KEM ciphertext + signature + AEAD) is stated to exceed 4.4 KB.
• Two envelope designs.
• Hybrid: a post-quantum KEM plus an obfuscated classical signature for real-time authentication. This is described as HNDL-resilient.
• Pure PQ: uniformized NTRU signatures (Falcon-512), or unilateral authenticated KEMs / signcryption.
• Named uniformization options. Power-of-two LWR lattices, with Saber and FrodoKEM given as examples, or constant-time inverse rejection sampling. The target statistical distance is Δ(P, U) ≤ 2^-λ.
• Why the distinguisher is cheap. With q = 3329, naive bit-packing leaves the upper sub-intervals unpopulated. The post claims Δ → 1 and an O(1) passive DPI check at line rate.
• Why fragmentation hurts. eBPF/XDP filters see individual frames, so a fragmented 0-RTT payload makes the responder allocate IP/UDP reassembly state before AEAD verification. The post says this reopens memory-exhaustion DoS and also creates multi-packet fingerprints.
• Silent-drop design. FO implicit rejection makes random probes and cross-node replays yield pseudorandom keys in constant time. They then fail AEAD verification and are dropped silently.

flowchart TD
    A[Classical tunnel moves to ML-KEM] --> B{Ciphertext on the wire}
    B -->|Bit-packed mod 3329| C[Non-uniform bytes: passive DPI distinguisher]
    B -->|Uniformized: LWR or rejection sampling| D[Looks uniform]
    A --> E{Envelope size}
    E -->|Over 4.4 KB| F[Fragmentation: reassembly state before AEAD]
    E -->|Within 1232 or 1472 bytes| G[Per-packet O1 drop in XDP]

Caveats

• This is a self-published preprint announced on a forum, with no peer review or empirical evidence in the post.
• The post lists FrodoKEM under "LWR lattices", but FrodoKEM is based on plain LWE with a power-of-two modulus. Check the preprint's exact claim before relying on that grouping.
• Whether Falcon-512 fits the sub-MTU budget alongside a KEM ciphertext, and any wire-format specifics, are not shown in the post.
• Relevance to Ethereum validator, libp2p and RPC transports is asserted rather than demonstrated. No concrete integration with any client is described.