Combating deepfakes and protecting digital integrity: Combining blockchains with verifiable off-chain compute
OpenAI just announced Sora, its text-to-video product, and it is blowing everyone’s mind.
Given how far text-to-video genAI has come in one year, it’s no surprise that forward-thinking technophiles are concerned about its implications wrt deepfakes and our ability to discern what is AI-generated vs. human-generated.
We see AI quickly ushering in a new era of rampant misinformation and deepfakes, posing a serious challenge that necessitates effective countermeasures. Here, we will explore how blockchains combined with verifiable off-chain compute architecture including zero-knowledge proofs (ZKPs), multi-party computation (MPC), and fully homomorphic encryption (FHE) present an innovative solution that may lay the groundwork for one of crypto’s next killer use cases.
Blockchains are the foundational infrastructure within this prospective ecosystem as they offer an immutable and transparent record of digital transactions that can verify content authenticity. They allow content to link with a verifiable cryptographic signature that can then be published to an openly readable, difficult-to-corrupt database.
However, while blockchains alone are not sufficient for handling sensitive data due to privacy and scalability issues. Personally harmful deepfakes (i.e., involved in defamation, fraud) need more than this, they need countermeasures that are verifiable and open and privacy-preserving.
Enter ZKPs. They enable a user to verify data including sender, receiver, and the content communicated without revealing any identifiable information about these, ensuring privacy and security. ZKPs work by allowing one party (the prover) to prove to another (the verifier) that a statement is true, without revealing any information beyond the fact that the statement is true. For example, a user could log their location or identity to a service without revealing their exact whereabouts or other PII.
Imagine a public figure accused of being in a controversial location at a specific time, based on a deepfake video. To refute this claim, the target could use a blockchain-based app that logs their location data, verified through ZKPs. The app confirms their presence elsewhere at the alleged time, without revealing their actual location to the public — a digital watermark effectively working as an alibi. The blockchain’s append-only design paired with ZKP-enabled privacy effectively can discredit false accusations.
Current deepfakes already allow any individual to create photorealistic representations of other people doing whatever they imagine.
Note, however, there is one major limitation to ZKPs that’s relevant in this context:
Due to computation demand, ZKP systems typically require users to trust third parties with more compute power to decrypt their data into plaintext to execute state transitions.
In other words, one must fully trust that third party — often a single centralized entity — not to expose or corrupt those plaintext data, which may include PII and other sensitive information.
Enter FHE + MPC
Technologies that innovate on the foundations of FHE + MPC could improve the overall security of any digital privacy scheme. For example, Nillion’s Nil Message Compute (NMC) enables off-chain computations on FHE-encrypted data fragments without the need for nodes to message each other, making for a more private and potentially more efficient solution. By adding an MPC architecture, users can be sure that only shards of their encrypted data are processed by any single node.
Combining ZKPs into this system can enhance overall privacy as they can verify the user’s identity or their access rights to certain data without disclosing their personal information. Sunscreen and Dock are working on this.
The practical applications are vast and extend beyond the individual. For example, in journalism, they could authenticate the provenance of news, such as its source and timing, providing a defense against fabricated stories. In elections, these technologies could safeguard the integrity of voting, verifying voter identities and proof of personhood without revealing any personally identifiable information, mitigating impersonation and other attempts at vote corruption. Relatedly, ensuring media integrity can protect against campaigns aimed at manipulating public perceptions around voting or any other social activity.
In sum, the convergence of blockchain with non-blockchain innovations like ZKPs, MPC, and FHE offer a powerful system for combating AI-generated misinformation and deepfakes. Blockchains lay the groundwork with immutable, difficult-to-corrupt public records in the looming era rife with digital fabrications. ZKPs bring a layer of privacy, enabling data verifications without personal data disclosure, while FHE and MPC tackle the issue of how to keep data securely encrypted from end to end.
If this post resonates with you or you’re building in these spaces, please do reach out! We would love to chat.
Disclaimer: This post is for general information purposes only. It does not constitute investment advice or a recommendation or solicitation to buy or sell any investment and should not be used in the evaluation of the merits of making any investment decision. It should not be relied upon for accounting, legal or tax advice, or investment recommendations. This post reflects the current opinions of the authors and is not made on behalf of Inception Capital or its affiliates and does not necessarily reflect the opinions of Inception Capital, its affiliates, or individuals associated with Inception Capital. The opinions reflected herein are subject to change without being updated.
