The Ledger
Andrey Chmora Proposes Stealth ABBE for Private Ethereum Contracts
Andrey Chmora’s Ethereum Magicians post outlines a BLS12-381 broadcast-encryption design for distributing private contracts and encrypted state. Its 24.1 KiB frame size is amortized, and a later revision adds a public salt to an open transport header.

Andrey Chmora’s November 4, 2026, Ethereum Magicians post proposes a BLS12-381 framework for distributing private smart contracts, confidential L2 or Validium state diffs, and permissioned intent payloads over Ethereum’s public P2P and Data Availability layers. The design, called Stealth Attribute-Based Broadcast Encryption (Stealth ABBE), aims to hide access policies and recipient information in the broadcast.
Chmora writes: “Deploying and distributing private smart contracts, confidential L2/Validium state diffs, or permissioned intent payloads across Ethereum’s public P2P and Data Availability (DA) layers currently faces a fundamental metadata bottleneck.” The post describes conventional broadcast-encryption and attribute-based encryption approaches as exposing recipient indices, covering subsets, or access trees in cleartext.
The proposed KEM frame size depends on the encryption setup
The post gives an O(1) broadcast KEM size of 24.1 KiB amortized, independent of the total number of participants. It says this fits within a single EIP-4844 blob, which the post describes as 128 KiB. The 24.1 KiB figure is for an amortized frame with a shared component; a frame using independent ephemeral scalars is 48.0 KiB.
For a private contract, the proposal symmetrically encrypts the bytecode and state, then posts a commitment—such as a code hash and state root—to an Ethereum registry contract. Its KEM payload is 192 bytes, containing the session key and cryptographic binding metadata. The payload is encoded as a non-systematic Reed–Solomon codeword with a 64-byte MDS puncturing defense margin, then masked with a policy token.
Access control is encoded into a Reed–Solomon code
Stealth ABBE frames access control as blind Reed–Solomon decoding over 𝔽₂₅₆. Participants hold locally stored ElGamal-encrypted access codes that encapsulate masked attribute vectors. The broadcast combines a policy token with the encoded session-key material. Under the proposal, only participants whose provisioned attributes satisfy a k-out-of-n threshold can recover the session key and decrypt the content.
The design uses a non-systematic Reed–Solomon code and describes access policies, cohort rosters, and revocation states as algebraically absorbed into the codeword. The initial post specifies zero visible headers and revocation lists. A subsequent revision adds a public salt in an open transport header, so the design is not accurately described as transmitting no metadata at all.
Base-3 packing and the outer transport layer
The proposal encrypts codeword blocks using additively homomorphic exponential ElGamal over 𝔾₁ × 𝔾₁. Because ciphertext addition in the exponent uses integer addition, binary 1 + 1 can produce a carry. The post says it avoids that problem by packing 8-bit blocks with Base-3 weights; the broadcast vector has 256 blocks.
For the outer transport, the post describes mapping points through constant-time inverse isogenies and Elligator Squared into pairs of field elements. It presents the resulting data as statistically indistinguishable from uniform channel noise. A reply disputes whether the serialized bytes are indistinguishable from uniform channel noise. The post’s 24.1 KiB frame figure should be read with its amortization condition, rather than as the size for independent ephemeral scalars.
Revocation requires more than Tier 1 for full eviction
Chmora later clarified that Tier-1 revocation is an operational quarantine: it cannot fully evict a node that retained information from an earlier frame. Full eviction requires Tier-2 rotation and new access codes. This qualifies the initial post’s claim that compromised nodes or auditors can be evicted instantaneously in O(1) without re-keying active participants.
The post also proposes using BLS12-381 through EIP-2537. It describes a decentralized federation using distributed key generation, with the global master secret never reconstructed in plaintext. These are elements of the proposed architecture, not a report that the system has been deployed.
Sources
- Original source Ethereum Magicians ethereum-magicians.org