FE Covenants Would Put Bitcoin Rules in the Key, Not Code
In this article
A research report published on 18 November 2026 examines a 2024 paper by Jeremy Rubin, the original author of the OP_CTV proposal under BIP-119, that proposes enforcing Bitcoin covenant behavior through functional encryption (FE) rather than new consensus rules. Rubin’s design, titled FE’d Up Covenants, places the spending restriction inside the signing key instead of Bitcoin Script. The analysis of that design, published by Cointelegraph, concludes that the cryptography it relies on remains impractical with current computing resources.
Changing Bitcoin’s consensus rules demands coordination among node operators, miners, developers and economic actors, a process the report describes as difficult to align, which is why the base layer has gradually ossified. Rubin’s approach sidesteps that coordination by moving the restriction off-chain, into cryptography that only the signer has to perform.
How the key does the enforcement
Ordinary public-key cryptography allows a private key to sign any message. Under Rubin’s construction, a setup authority creates a Bitcoin key pair and a functional encryption key pair, then publishes the public key, the FE public key, an encrypted copy of the private key and a functional key for a signing function. The authority must be trusted to destroy both master secrets, since anyone holding them could sign spends that break the covenant.
Each covenant then gets its own key pair, derived by tweaking the base key with a hash of the spending policy. The private half exists only inside the signing function and is never exposed, while the public half can be derived by anyone who knows the policy hash, which is how receiving addresses can be generated. In practice, the report says, a receiving address would commit to a 2-of-2 script requiring a user-held key alongside the FE signature, to stop unauthorized parties from triggering a permitted spend.
To spend, a signer encrypts a transaction that conforms to the policy, evaluates the signing function and receives a signature tied to the tweaked key. A transaction that violates the policy could at best be signed for some other, differently tweaked key that does not control the funds. Rubin’s general form, f(k, tx, program, proof), is strictly more expressive than OP_CTV because it takes a program plus a zero-knowledge proof that the program accepted the transaction, and it can even depend on data outside the transaction.
Why the scheme is not deployable yet
Multi-input FE, introduced by Goldwasser and co-authors in 2014, was built on indistinguishability obfuscation, and the two primitives imply one another. Ananth and Jain later showed that compact single-input FE for general circuits also yields obfuscation. The primitive therefore inherits the difficulty of building obfuscation itself, and existing implementations stay at toy scale.
The report cites Carmer, Malozemoff and Raykova, who in 2017 spent 3.7 hours obfuscating a 12-bit pseudorandom function: a circuit of roughly 62,000 gates with a 64-bit key, far smaller than ECDSA signing on 256-bit numbers. The obfuscated program consumed 120 GB of disk, and one evaluation took 67 minutes with 128 GB of RAM. The authors described the result as still far from practical, and noted the cost would fall on whoever produces the signature rather than on Bitcoin nodes. Practical FE schemes cover only narrow function classes, and ECDSA signing is not one of them.
Why it matters
The comparison sits at the center of the report: opcode-based designs closer to OP_CTV remain far more efficient in bytes and verification costs but need a base-layer change, while FE covenants avoid consensus changes yet depend on cryptography nobody can yet run. For Bitcoin holders and developers, that means the covenant debate is unlikely to be settled by off-chain cleverness alone any time soon. Those following related cryptography may recall BitcoinWorld’s earlier coverage of UK banks blocking third-party AI agents from customer accounts, another case where trust assumptions and enforceability determine whether a design reaches production.
What to watch
The series continues with Un-FE’d Covenants, Rubin’s proposal to approximate covenant behavior using oracle signers bonded through BitVM, followed by a look at a proposal that began in 2024 as an FE design and moved to witness encryption this year.
Market snapshot
Bitcoin traded at $62,475 (₹5,837,883) on 18 November 2026 at 20:48 UTC, down 2.9% over 24 hours and 16.1% over seven days, with a market cap of $1.26 trillion and $19.60 billion in 24-hour volume, according to CoinPaprika data.
Frequently Asked Questions
What is a functional encryption covenant?
It restricts a signing key so that it can only produce a valid signature for transactions that match a predefined spending policy, instead of requiring new opcodes on Bitcoin’s base layer.
Why do FE covenants avoid a soft fork?
Bitcoin nodes would still verify signatures through OP_CHECKSIG as they do today, so no change to consensus rules is required.
Why aren’t FE covenants usable now?
Multi-input functional encryption implies indistinguishability obfuscation, and the only known implementations remain at toy scale with costs far beyond practical use.
Who proposed FE covenants?
Jeremy Rubin, the original author of the OP_CTV proposal under BIP-119, presented the design in a 2024 paper titled FE’d Up Covenants.
What comes next in the research series?
A follow-up part will cover Rubin’s Un-FE’d Covenants and a separate proposal on witness encryption.
Source: Cointelegraph



