bitpro가 원문에서 글자만 모은 내용이에요. 표·그림·수식은 빠지거나 깨질 수 있어요. 전체 45쪽 중 1쪽
Whitepaper (2024)
Interfold: A Protocol for Encrypted Execution Environments (E3)
Auryn Macmillan
Nathan Ginnever
Marvin Lin
Note:
This whitepaper describes the original Interfold protocol design.
Interfold has since evolved into The Interfold. This document is preserved here as an archival
technical reference.
Abstract
This whitepaper introduces Interfold, an open-source protocol for Encrypted Execution Environments
(E3). By integrating fully homomorphic encryption (FHE) with a novel combination of zero-knowledge
proofs (ZKPs) and distributed threshold cryptography (DTC), Interfold enables secure computations on
multiple private inputs using a decentralised network of nodes that decrypts only the verified
outputs from the requested FHE program. This privacy-preserving framework relies on a
stakeholder-driven consensus mechanism, providing strong economic and cryptographic guarantees
around data privacy, data integrity, and correct execution.
Interfold's trust assurances enable a new class of applications characterised by liability-free
access, collaborative data analysis, and provably legitimate outcomes. The latter use case is
demonstrated through Interfold’s flagship application CRISP (Coercion-Resistant Impartial Selection
Protocol), a secret ballot protocol aimed at preventing coercion and mitigating vulnerabilities in
digital decision-making. The paper concludes by exploring how Interfold can be implemented in other
sectors, from finance and healthcare to AI and beyond.
Introduction
In our increasingly data-driven environments, the lack of robust data security and computational
integrity poses significant challenges. Organisations risk privacy breaches or legal liabilities
when leveraging sensitive data, while applications that require guarantees around data privacy and
correct execution are either constrained by trust or simply non-existent. Current solutions are
traditionally dependent on trusted operators or trusted hardware (begging the question of
trust-dependent supply chains), introducing untenable counterparty risk for many applications. This
can force a trade-off between data utility and regulatory compliance, leaving valuable data either
vulnerable, underutilised, or unused.
Interfold addresses these challenges with Encrypted Execution Environments (E3), enabling secure
computations on multiple private inputs secured by a decentralised network of nodes. By integrating
fully homomorphic encryption (FHE) with a novel combination of zero-knowledge proofs (ZKPs) and
distributed threshold cryptography (DTC), Interfold’s nodes are designed to decrypt only the
verified outputs from the requested FHE program, providing strong cryptographic guarantees around
data privacy, data integrity, and correct execution. This computational infrastructure allows
organisations to both safeguard and leverage sensitive data, opening new opportunities for
innovation that preserve privacy and ensure compliance.
E3 is designed to handle any use case where it is valuable to aggregate or compute over multiple
inputs from various sources, guaranteeing that inputs and intermediate states remain private,
computations are secure, and outputs are verifiable. Due to Interfold's modular and composable
design, E3 is compatible with a wide variety of encryption schemes and can securely execute
arbitrary computations over encrypted data. The protocol's ephemeral execution further minimises
risks, while single-use keys eliminate the underlying vulnerabilities and economic inefficiencies
that exist with global keys controlled by a single party. While leveraging blockchain technology to
inherit some of its desirable properties (censorship resistance, verifiability, and data
availability), Interfold’s infrastructure is broadly applicable both within and beyond the web3
ecosystem.
Interfold’s use cases can be categorised into three main classes:
Provably Legitimate Outcomes: Interfold enhances the legitimacy of outcomes by eliminating the
need for a trusted operator (e.g. an auctioneer) or trusted hardware (Trusted Execution
Environment) to compute over multiple private inputs. Examples include secret ballots and
sealed-bid auctions.
Liability-Free Access: Interfold facilitates secure computations over sensitive data using FHE,
reducing liability and increasing data accessibility. This is critical for applications like
medical research and financial data analysis, where privacy and compliance are critical.
Collaborative Data Analysis: Interfold allows multiple entities to collectively compute on
encrypted data without sharing raw datasets, preserving the privacy of sensitive or proprietary
information. This innovation enables fields like genomics and climate science to securely
collaborate on data while maintaining full confidentiality of their inputs.
At the heart of Interfold's operational model is a decentralised network of staked nodes responsible
for overseeing data processing. Through a multiphase process, these data-processing interfolds
ensure the correct execution of computations and the confidentiality of private data through a
consensus mechanism driven by stakeholder participation. The strategic use of proportional fees,
punitive measures, and staking mechanisms increase transparency and accountability, pairing strong
economic guarantees with cryptographic guarantees around privacy, liveness, and integrity.
In this whitepaper, we delve into Interfold's core principles, technical details, and
functionalities, highlighting flagship application CRISP (Coercion-Resistant Impartial Selection
Protocol), which utilises E3 to enable verifiable secret ballots. By exploring the design of CRISP,
we demonstrate how Interfold’s modular framework can be adapted to various use cases and needs,
maximising the value of sensitive data while enhancing privacy, integrity, and trust.
Overview
Interfold is an open-source protocol for coordinating the various actors and composing the various
components necessary to create Encrypted Execution Environments (E3). These novel execution
environments secure computations using fully homomorphic encryption (FHE) and other advanced
cryptographic techniques, like zero-knowledge proofs (ZKPs) and distributed threshold cryptography
(DTC). Unlike Trusted Execution Environments (TEEs), which rely on trusted hardware and can be
vulnerable to side-channel attacks and validation weaknesses, Interfold's decentralised
architecture, advanced cryptography, and privacy-preserving mechanism design provide strong economic
and cryptographic guarantees around data privacy, data integrity, and correct execution.
The efficacy of these guarantees stems largely from Interfold’s distinct modular architecture, which
allows for adaptable and resilient computation across a variety of use cases. By leveraging a
network of staked nodes that participate in a multiphase computational process, Interfold integrates
a sophisticated economic model featuring staking mechanisms, proportional fees, and punitive
measures to incentivize correct behaviour. This design not only mitigates the risks associated with
trusted systems but also provides a versatile foundation for building privacy-preserving
applications tailored to various sectors, ensuring that organisations can leverage sensitive data
without compromising security or compliance.
Actors in Interfold
Interfold enables secure, privacy-preserving computations on encrypted data through the coordination
of three distinct Actors within the protocol. The following is a description of each role and its
contribution to the process:
1. Requesters
Requesters initiate the process and define the parameters for secure computation. Any individual or
organisation can request an E3 from the Interfold network by calling the corresponding smart
contract entrypoint and depositing a bond proportional to the number of nodes, threshold levels, and
duration of the E3 requested.
2. Data Providers
Data Providers are individuals or systems that supply the data inputs for a requested E3. Their
plaintext data is encrypted using a public threshold key generated by and published on chain by the
requested E3, ensuring the data remains secure and private throughout the process.
3. Encrypted Execution Environments (E3)
E3s are the core units that execute secure computations on encrypted data, requested from and
coordinated via the Interfold protocol. E3s use a decentralised network of nodes to decrypt only the
verified outputs from the requested FHE program. This guarantees that computations are performed
while keeping data private and secure.
E3 Architecture
E3s consist of several modular components that ensure secure and verifiable computations on
encrypted data. These components are designed to maintain data privacy and computational integrity,
mitigating risks associated with trusted solutions.
Each E3 is composed of the following key components:
Ciphernode Committee (CiCo)
A selected group of Ciphernodes responsible for both distributed key generation and decrypting the
ciphertext output published by the Compute Provider. Nodes are selected for committees using a
randomised sortition, enhancing security by distributing trust. This decentralised approach ensures
that no single node has complete control, eliminating the risk of data breaches or tampering.
E3 Program (E3P)
The specific set of instructions that defines the computation to be performed. The E3 Program takes
ciphertexts published onchain as inputs and produces plaintext outputs, ensuring the confidentiality
of both data and results. E3Ps are designed to be versatile, supporting various computations while
maintaining security.
Secure Process (SP)
An E3P component dedicated to running FHE computations and executed by the Compute Provider. Unlike
other system components managed by the protocol, Ciphernodes, or Data Providers, the Secure Process
runs exclusively on the verifiable component of the Compute Provider, which delivers onchain proof
or validation of an SP’s correct execution.
Compute Provider (CP)
The platform executing the Secure Process in an E3. The Compute Provider is implicitly trusted to
report back the correct ciphertext outputs as a result of the Secure Process running over the
published inputs for the given E3. CPs can be categorised into proof-based and oracle-based systems:
Proof-based providers (RISC Zero
1
, Succinct’s SP1
2
, Jolt
3
, Nexus
4
) offer cryptographic
guarantees of correct execution, while oracle-based providers (Multisigs, Reality.eth
5
, Kleros
Court
6
, API3
7
) use consensus or economic mechanisms for validation. This modularity allows E3s
to select the most appropriate CP based on the specific cost, security, and trust requirements.
How Interfold Operates
Interfold facilitates secure, decentralised computation through a structured multiphase process
involving the various Actors and E3 components. This process guarantees correct execution while
safeguarding privacy. Figure 1 is a visual depiction of these interactions.
Phase 1: Request
Anyone can initiate a computation request from Interfold at any time. This request must specify the
following:
The FHE computation to be performed.
The choice of Compute Provider and associated parameters.
The number of Ciphernodes required.
The threshold for each type of node needed for output validation.
The timestamp for Input Deadline, after which no new inputs will be accepted.
The duration for which the nodes must be available.
To initiate a request, a deposit proportional to ensure a minimum reward for the requested nodes for
performing their duties is required. The deposit amount can be increased by other participants up
until the final phase, Decryption.
Phase 2: Node Selection
When a computation is requested, the required number of Ciphernodes are selected to form a
Ciphernode Committee through a random process called sortition. The selected Ciphernodes immediately
generate and publish shared public keys with the requested thresholds. This securely establishes the
cryptographic framework for the computation.
Phase 3: Input Window
Once the selected Ciphernode Committee has published their shared public key, the Input Window
opens. During this phase, Data Providers can submit their inputs for the computation, which are
encrypted to the public key to ensure confidentiality. Each input requires a corresponding
zero-knowledge proof (ZKP) to ensure the Data Provider knows the preimage to their input, as well as
that the input is correctly formed, cannot be used in any other context, and passes any other
validation logic required by the selected computation.
Phase 4: Execution
After the Input Window closes, the selected Compute Provider executes the predefined E3 Program,
including the Secure Process that runs the FHE computation, using the encrypted inputs. The Compute
Provider then generates and publishes ciphertext outputs, ensuring that the computation's integrity
remains intact throughout the process.
Phase 5: Decryption
Once the selected Compute Provider publishes the ciphertext outputs of the requested computation, a
threshold of the Ciphernodes must collectively decrypt and publish the output of the computation as
reported by the Compute Provider. After the plaintext of the computation output is published,
Ciphernode duties are complete. Each node can then claim its proportional share of the bond and
should dispose of the keys used for this committee, treating them as toxic waste.
Incentives and Penalties
To maintain integrity and encourage correct behaviour, Interfold utilises a system of incentives and
penalties.
Staking
Anyone can register a new Ciphernode by staking an amount
8
of Interfold tokens. After
registration, new nodes must wait for a regitration delay period
9
before they can be selected for
Ciphernode duties. The registration delay period enables Requesters to reasonably predict the
current Ciphernode set when requesting a computation. This economic stake aligns their interests
with the security of the system.
Nodes can request to be decommissioned at any time. Prior to being decommissioned, nodes must remain
active for a decommission delay period
10
, after which they will no longer be selected for duties.
Nodes must also remain active to complete any duties for which they have been selected. Once its
decommission delay has passed and all assigned duties have been completed, a node may be
decommissioned, and its staking deposit will be returned.
Fees
When requesting a computation from Interfold, Requesters must deposit a fee proportional to the
number and threshold of Ciphernodes requested, the duration for which they are required to be
available, and any additional costs required by the selected Compute Provider. This deposit reserves
the requested Ciphernodes, providing an economic guarantee that they will be online to decrypt the
output of the requested computation. However, the deposit does not pay for the computation itself.
Different Compute Providers will have different fee structures which are levied outside of the
Interfold protocol.
Ciphernodes are subject to a penalty and forcefully decommissioned if they are proven to have
provided their share of the decryption data from any input or intermediate states for a computation
on which they were a committee member. To ensure there is a long-term disincentive for intermediate
decryption, even for decommissioned Ciphernodes, a portion of the Ciphernode rewards are paid out
immediately after Decryption, with the remaining portion subject to a cliff and vesting
schedule
11
, along with slashing conditions.
Penalties
Malicious behaviour or failure to adhere to protocol rules results in penalties, ranging from loss
of rewards for a given duty round to forced decommissioning. This disincentivizes actions that could
compromise data security or computational integrity. The cases and penalties are detailed here:
Inactive Node:
Ciphernodes that fail to provide their share of the data necessary for
decrypting a requested computation forfeit their share of the decryption fee. They are also
subject to a small penalty
12
used to cover the additional gas cost incurred due to their
missing signature.
Failure To Report Outcome:
If the selected Compute Provider does not provide the ciphertext
output of the requested computation in a timely manner
13
, the computation request is cancelled.
In this case, a small portion of the bond
14
is paid out to the selected Ciphernode Committee,
while the remainder of the bond for the computation is returned to the Requester.
Failure To Decrypt:
If the selected Ciphernode Committee does not provide the decrypted
plaintext from the ciphertext output agreed on by the selected Compute Provider in a timely
manner
15
, the computation request is cancelled. In this case, the bond for the computation is
returned to the Requester. Any nodes not providing their share of the decryption data are subject
to a penalty and are forcefully decommissioned. Penalties collected are split proportionally
between the remaining members of the selected Ciphernode Committee.
Non-Deterministic Computation:
In order for nodes to reach consensus on the output, a
requested computation must be deterministic
16
. Suppose the selected Compute Provider reports
that the requested computation has a non-deterministic output. In that case, the bond for that
output is partially distributed between the selected nodes and partially used to purchase and burn
tokens.
Intermediate Decryption:
Maintaining the privacy of all encrypted inputs and intermediary
states is a critical feature of Interfold. Attempts to decrypt anything encrypted to a Ciphernode
Committee, except the agreed upon output of a requested computation, is punishable by forced
decommissioning, along with slashing
17
a portion of the offending node’s stake. A portion of
the slashed stake is burned, while the remaining portion is allocated to the account that notified
the network of the slashable offence.
Forced Decommissioning:
If a node’s effective stake is ever reduced to half of the minimum
stake, it is immediately decommissioned and will no longer be selected for duties. However, nodes
must also remain active to complete any duties for which they had already been selected. Once all
assigned duties have been completed, a forcefully decommissioned node may claim the remainder of
its staking deposit.
Interfold Governance
Interfold is governed by its token holders, who serve as the top-level decision-making body. They
are responsible for setting protocol parameters, overseeing protocol upgrades, and facilitating
dispute resolution, among other duties. Interfold has several variables that may require periodic
adjustments to maintain the protocol's fairness, performance, and responsiveness to evolving
requirements.
The responsibilities of Interfold token holders include:
Protocol Management
Protocol Upgrades
Dispute Resolution
Parameter Settings
Cancellation Burn Percentage
Staking Deposit Amounts
Registration Delay Period
Decommission Delay Period
Inactive Node Penalty Amounts
Intermediate Slashing Penalty and Burn Ratio
Flagship Use Case: Secret Ballots
Contemporary decision-making systems are often compromised by issues such as coercion, data
breaches, and privacy violations. These vulnerabilities undermine the integrity and effectiveness of
governance and decision-making systems, creating a need for a secure and reliable solution.
The Coercion-Resistant Impartial Selection Protocol (CRISP) addresses these challenges by ensuring
decision-making systems remain secure and equitable through Interfold’s advanced cryptography.
Serving as a modern embodiment of the secret ballot, CRISP is a secure protocol for digital
decision-making, leveraging FHE and threshold cryptography to enable verifiable secret ballots; a
critical component for democracies and many other decisio프로젝트로 돌아가기