Interactive simulation. Mock data, no functional backend, no deployable code. Built to study detection, not to operate.

The C2 framework that rewrites the rules of relay trust.

For fifteen years, every serious framework has made the same assumption: relay nodes are trusted. Novich0k breaks that assumption cryptographically. Relays transform traffic they cannot read, guaranteed by proof rather than policy, so detection engineering can evolve before it arrives.

See how PRE works Run the simulation
2-of-3 threshold PRE Per-message forward secrecy Verified live · edge deployment, April 2026 The break ↓

The break

The assumption was never challenged because the cryptographic tools to challenge it did not exist in deployable form. They do now.

A redirector was never protection; it was separation. In every traditional architecture, the relay terminates TLS, holds session keys, or sits one routing step from the team server. Seize it through abuse reports, legal process, or compromise, and it becomes a roadmap: the traffic, where it went, the team server. That is not a bug in Cobalt Strike, Havoc, Sliver, Brute Ratel, or Mythic. It is a structural property of how they were designed.

Novich0k replaces trust with mathematics. Its relays are edge workers holding key fragments (kfrags) of a threshold proxy re-encryption scheme. A worker transforms ciphertext from one key domain to another; it never decrypts, never touches plaintext, never holds a session key. Compromise it completely, down to every KV entry and every intercepted request, and the yield is still kfrags and ciphertexts. Fragments cannot decrypt; re-encryption in one direction yields nothing about the other. That is not configuration. It is what the construction allows, under the same formal proofs that underpin modern public-key cryptography.

Why publish it. A relay that cannot read its own traffic changes what takedown, seizure, and content inspection are worth. Frontier Labs models that shift in the open, pairing every capability with the detection surface it leaves behind, so defenders inherit the map first.

The mechanism, precisely

One diagram, four terms. This is the entire trick, and every property the project claims falls out of it.

Fig. H1 · a command's path: transformed in transit, readable only at the ends
Command path through a blind relay The operator encrypts under their own key and signs. The worker re-encrypts the capsule with a key fragment, producing capsule fragments, without decrypting. The beacon verifies the signature and fragments, then decrypts with its secret key. OPERATOR encrypts under op_pk + ECDSA signature WORKER kfrag × capsule → cfrags never decrypts BEACON verifies sig + cfrags decrypts with beacon_sk capsule + ciphertext ciphertext + cfrags The payload never changes in transit. Only the capsule, the mathematical envelope, is transformed.
Text equivalent
  1. The operator encrypts the command under their own public key and signs the envelope.
  2. The worker applies a key fragment to the capsule, producing capsule fragments. It performs a transform, not a decryption; it never sees plaintext.
  3. The beacon receives the same ciphertext plus the fragments, verifies the operator's signature and each fragment's correctness proof, then decrypts with its secret key.
TermWhat it isWhy it matters
CapsuleThe PRE ciphertext header, carrying a fresh ephemeral key generated per messageForward secrecy at message granularity: later key compromise opens nothing already sent
kfragOne threshold share of a re-encryption delegation, the only key material a relay ever holdsBelow threshold, fragments are information-theoretically empty; seizure yields nothing
cfragThe relay's re-encryption output, carrying a cryptographic proof it was produced honestlyThe endpoint verifies the relay behaved correctly without trusting it
2-of-3 thresholdThree fragments issued per direction; any two suffice, one reveals nothingPartial takedown degrades nothing; no single provider can act or inform alone

CONFIRMED: the full pipeline (registration, command, result, signature and replay rejection, failover) was verified against a live Cloudflare edge deployment in April 2026. Full architecture →

The shift, in three moves

Takedown economics assume relays can read traffic and that seizing one tells you something. Each move below removes one of those assumptions.

Today: trusted relay operator beacon relay plaintext one seizure reads everything

One server holds the command queue, session keys, and history. Seize it, subpoena it, or watch it - the whole operation is exposed.

Move one: blind relay operator beacon relay ciphertext seized relay = unreadable blobs

Proxy re-encryption lets the relay transform messages it can never read. Seizure yields ciphertext, indistinguishable from random data.

Move two: threshold blind relay k₁ k₂ k₃ k₄ k₅ seized any 3 of 5 suffice

The re-encryption key is split across a pool. Any m of n fragments suffice; fewer reveal nothing. No single seizure suffices.

What the research establishes

Five claims the Novich0k track defends. Each runs in the built system, and each is documented in the write-ups.

  1. PRE as a C2 transport. Proxy re-encryption has lived in data-sharing and key-management literature. This is its first application as the transport layer of command infrastructure.
  2. Blindness by mathematics, not configuration. Relay blindness is a property of the Umbral threshold scheme on secp256k1. It cannot be undermined by a misconfigured or compromised relay.
  3. Zero runtime key exchange. All key material is generated offline. The beacon's first network contact is a data poll, eliminating key establishment as an attack surface and as a detection fingerprint.
  4. Threshold relays as infrastructure. The m-of-n fragment scheme is applied to relay nodes themselves, so partial takedown degrades nothing.
  5. Re-encryption at the edge. Genuine PRE executed in serverless edge workers via WASM: a relay that deploys in seconds and burns without forensic residue.

Why now

The building blocks went mainstream. Zero-trust architecture and confidential computing are now standard defensive practice, and attackers adopt proven patterns. The cryptography that removes trust from relays is published, audited, and a library call away.

Relay capacity became disposable. Commodity edge workers and key-value storage make a relay pool trivially cheap to run and nearly free to replace. A seized relay is no longer a loss; it is a line item.

Detection economics assume relay visibility. Takedown playbooks, seizure warrants, and relay-level monitoring all price in a relay that can read. When the relay is blind and disposable, that price is wrong.

This is pathfinding research: model the threat before it is the threat.

Research tracks

Each track reads on its own. Together they trace one argument: problem, mechanism, consequence, demonstration, response.

PRE Primer

Proxy re-encryption from first principles, diagram-led, no cryptography background assumed.

Learn the mechanism

Architecture

How Novich0k removes the relay from the trust equation, and what that inversion costs defenders.

Read the architecture

Traffic Research

Individually irregular, collectively correlated. Compare conventional, legitimate, and research-model traffic.

See what survives

Adversarial Simulation

Attack it or defend against it. Two independent research simulations over the same Novich0k model.

Choose a side

Detection Surfaces

Blind relays remove content, not physics. The residual signals, hunting hypotheses, and open questions.

Read the response

VileBlood

Process fragmentation for endpoint research: one implant's duties split across three isolated processes.

See the model

Publications

Long-form writeups, methodology notes, and the author's contact - the citable record of the work.

Read the articles

Research ethics

All research is conducted in closed lab environments for defensive and academic purposes. Nothing on this site is deployable, and every capability described is paired with what defenders can still see. Read the methodology →