Jeremy Rubin's 2024 Paper Ties Bitcoin (BTC) Covenants to Signing Keys
A 2024 design by Jeremy Rubin would enforce Bitcoin (BTC) spending rules without a soft fork, but the cryptography it needs remains impractical.
AI SummaryAI
- Jeremy Rubin proposed FE-based Bitcoin covenants in a May 2024 paper titled FE'd Up Covenants.
- Functional encryption lets nodes verify signatures with OP_CHECKSIG, avoiding any soft fork.
- A setup authority must generate and then destroy the private key and FE master secret.
- A 2017 experiment obfuscated a 62,227-gate circuit in 3.7 hours using 120 GB of disk.
Covenants in the Signing Key
Covenant research that sidesteps a soft fork returned to center stage on May 29, 2024, when Jeremy Rubin, the author of the OP_CTV proposal registered as BIP-119, published a paper titled FE’d Up Covenants. The design moves the spending rules of
Bitcoin (BTC) out of Script and into the signing key itself, so nodes never need new consensus code. The work targets the protocol layer rather than market activity, and the Bitcoin price is not directly affected.
The idea answers a structural constraint. Changing the base layer requires node operators, miners, developers and economic actors to coordinate, and that difficulty is why the Bitcoin protocol has ossified, a stance long associated with Bitcoin Maximalism. A covenant restricts what an output can be spent on; ordinary signature verification cannot impose such a limit, which is why the proposal space keeps growing. Field mapping has catalogued more than 24 covenant use cases, ranging from vaults to contract-state chaining under the OP_CCV draft (BIP 443) and into design territory shared with Bitcoin DeFi, Bitcoin Ordinals and Bitcoin Runes.
Mechanically, functional encryption generalizes public-key cryptography. Instead of one private key able to sign anything, a master secret issues subsidiary functional keys, each revealing a single chosen function of a plaintext, a construction first formalized by Dan Boneh, Amit Sahai and Brent Waters in 2011. For covenants, that function is a signer: it takes an encrypted private key and an encrypted transaction, derives a fresh key pair tweaked by a hash of the spending policy, and returns a signature valid only for the tweaked public key. A transaction that violates the policy could at best be signed for some other, differently tweaked key that controls nothing. Nodes verify the result with OP_CHECKSIG exactly as they do today, and covenant addresses commit to a 2-of-2 script pairing the user’s key with the covenant signature, keeping outsiders from triggering a permitted spend.
A 62,000-Gate Toy Benchmark
The costs land on whoever produces the signature, not on nodes, and the primitive the design needs is at least as hard to build as indistinguishability obfuscation. Multi-input functional encryption was introduced in 2014, and Ananth and Jain showed in 2015 that even compact single-input FE for general circuits yields obfuscation. Implementations of that principle remain toy scale. A 2017 experiment by Carmer, Malozemoff and Raykova obfuscated a 12-bit pseudorandom function, a circuit of about 62,227 gates with a 64-bit key, in 3.7 hours; the obfuscated program occupied 120 GB of disk, and one evaluation ran 67 minutes with 128 GB of RAM. The authors called the result still far from practical, and the circuit is far smaller than ECDSA signing on 256-bit numbers.
Practical FE schemes cover only narrow function classes, such as inner products, and the ECDSA signing that
Bitcoin (BTC) covenants require is not among them. Rubin’s paper concedes the point and lists under-developed cryptography plus the setup trust assumption among its downsides. The trust issue is structural: a setup authority generates the Bitcoin key pair and the FE master secret, publishes the public key, the FE public key and the functional key, and must then destroy the private key and the master secret. If either survives, the authority can sign spends that break the covenant.
The direct alternative, a CTV-style opcode, is more efficient in bytes and verification but needs a base-layer change, and BIP-119 still carries Draft status. Rubin’s general construction pairs a condition-checking program with a zero-knowledge proof, making it strictly more expressive than OP_CTV and letting policies depend on data outside the transaction. A follow-up paper, Un-FE’d Covenants, replaces FE with oracle signers bonded through BitVM, and the Bitcoin PIPEs proposal, which began in 2024 as an FE design, moved to witness encryption in 2026.
The Trust Trade-Off
The FE route exists because the base layer has effectively stopped moving, and its obstacle is not only mathematics. Covenants without a soft fork trade consensus risk for trust in a setup authority, a bargain that mirrors the ossification problem rather than dissolving it. The 62,227-gate benchmark sizes the gap: a production covenant signer would need circuits far beyond anything demonstrated, while every practical FE scheme to date covers narrow function classes only. That leaves the field’s near-term weight on two visible tracks, whether BIP-119 advances beyond Draft and whether BitVM-bonded oracle signers can approximate covenant behavior without the FE burden, and a follow-up installment covers the latter design.
AI-generated, AI-reviewed, under COINOTAG editorial oversight.

