Agora: A Privacy-Aware Data Marketplace

Vlasis Koutsos, Dimitrios Papadopoulos, Dimitris Chatzopoulos, Sasu Tarkoma, Pan HuiIACR ePrint 2020 · DOI

TL;DR

This paper introduces Agora, a novel data marketplace built on blockchain technology that enables secure data sharing and compensation for data generators and brokers. It leverages a new functional encryption scheme and zero-knowledge proofs to ensure data privacy, output verifiability, and atomic payments without relying on trusted third parties.

What Problem Does This Paper Address?

The increasing volume of data generated by internet-connected devices has led to the emergence of data marketplaces. However, existing solutions often require users to place significant trust in intermediaries (data brokers) to handle their sensitive data and ensure the accuracy of processed results. This raises concerns about data privacy, the potential for data manipulation, and the security of payment transactions. The paper addresses the challenge of creating a decentralized data marketplace that guarantees data privacy, allows consumers to verify the correctness of the data outputs they purchase, and ensures that payments are processed atomically, all without relying on trusted third parties or secure hardware.

Methods

Agora employs a multi-faceted approach combining several cryptographic techniques and blockchain technology:

  • Functional Encryption (FE): At its core, Agora uses a novel multi-client functional encryption (MCFE) scheme. This scheme allows data generators to encrypt their data such that only specific, predetermined functions of the data can be decrypted by authorized parties (brokers). This ensures that brokers only learn the output of the agreed-upon function, not the raw data itself. A key innovation is the introduction of a functional public key (fpk) that acts as a "trust anchor" for verifying decrypted results.
  • Zero-Knowledge Proofs (ZKPs): To address output verifiability, Agora integrates a tailored zero-knowledge proof protocol. This allows brokers to prove to consumers that the decrypted output they provide is indeed the correct result of the agreed-upon function applied to the encrypted data, without revealing any additional information about the data or the decryption key.
  • Blockchain and Smart Contracts: The entire system is built on a blockchain (specifically, Ethereum) that supports smart contracts. This provides a decentralized and transparent platform for managing transactions. Smart contracts are used to automate payments, ensuring atomicity – meaning either the payment and data exchange both occur, or neither does.
  • Batch Purchases: To improve efficiency and reduce transaction costs for consumers purchasing multiple data outputs, Agora implements a batch purchase optimization. This allows multiple verifications to be processed in a single on-chain transaction, making the cost constant regardless of the number of items purchased.

Datasets

Not discussed in the paper. The paper focuses on a general framework for data marketplaces and does not evaluate on specific public datasets.

Key Insights and Ideas

  • Decentralized Trust: The most significant insight is the ability to build a secure data marketplace without relying on a central trusted authority. By combining FE, ZKPs, and blockchain, Agora shifts trust from intermediaries to cryptographic guarantees.
  • Functional Encryption as a Privacy Enabler: The paper highlights how FE, particularly with the addition of the fpk, can provide granular control over data access, allowing only specific computations to be performed and their results to be revealed.
  • ZKP for Verifiable Transactions: The integration of ZKPs is crucial for building consumer trust. It allows consumers to independently verify the integrity of the data outputs they receive, even from a potentially untrusted broker.
  • Blockchain for Atomic Payments: Using smart contracts on a blockchain is an elegant solution for ensuring that payments are atomic, preventing scenarios where a consumer pays but doesn't receive the data, or vice-versa.
  • Efficiency through Batching: The optimization for batch purchases demonstrates a practical consideration for real-world applications, significantly reducing the cost burden on consumers.

Key Results

  • Prototype Implementation: A working prototype of Agora was implemented on Ethereum.
  • Performance: Experimental evaluations showed that Agora scales well with the number of generators. For instance, processing data from 1,000 generators for a weighted sum computation incurs a gas cost of approximately $31.62 USD, with decryption taking less than 0.5 seconds.
  • Cost Efficiency: The batch purchase optimization drastically reduces verification costs. For 10,000 outputs, the naive approach would cost $175.5 USD, while the optimized batch purchase reduces this to approximately $0.02 USD.
  • Security: The proposed MCFE scheme is proven secure against both active and passive adversaries under standard assumptions.

Limitations and Future Work

  • Limited Functionality: Currently, Agora primarily supports weighted sum computations. While expressive, this limits its applicability for more complex machine learning tasks.
  • Message Domain Size: The current implementation's decryption efficiency is tied to the size of the message space, which is limited by discrete logarithm computations.
  • Dynamic Generators: The system does not currently support dynamic changes in the group of generators.
  • Future Work: The authors suggest extending Agora to support more complex functions, such as quadratic functions, and exploring alternative FE and ZKP schemes to improve performance and expand functionality. They also note that the trusted authority for key generation could potentially be replaced by a decentralized protocol.

Summary updated

BibTeX
@article{koutsos2020agora,
  author = {Vlasis Koutsos and Dimitrios Papadopoulos and Dimitris Chatzopoulos and Sasu Tarkoma and Pan Hui},
  title = {{Agora: A Privacy-Aware Data Marketplace}},
  journal = {IACR Cryptol. ePrint Arch.},
  year = {2020},
  doi = {10.1109/icdcs47774.2020.00156},
}
Figure