Stateful ZK Identity with Ian Miers

Anna Rose welcomes Ian Miers back to the Zero Knowledge podcast to discuss his research on stateful zero-knowledge (ZK) identity, focusing on integrity, reputation, and identity in decentralized social networks. Ian's projects aim to develop tools for creating forums that can ban bad actors while preserving user anonymity. He emphasizes the importance of self-sovereign identities that evolve over time, ensuring they cannot be forged or reset.

Ian connects his work to real-world challenges such as account theft, trolling, and age verification, highlighting the need for maintaining data integrity in anonymous online interactions. He introduces the concept of ambient identity, suggesting that long-standing accounts could help manage online interactions without linking comments to a single revocable ID. Ian proposes ZK promises as a solution to ensure accountability in anonymous settings, advocating for civil liberties and free speech without tying all online comments to one identity.

The conversation explores the potential of ZK promises to create anonymous user identities while enforcing behavior-based restrictions. Ian discusses the evolution of identity and reputation systems, originally designed to combat money laundering, now addressing broader issues like social media bans. He emphasizes the need for these systems to reflect dynamic interactions rather than static attributes, highlighting the importance of programming rules that can range from simple bans to complex mechanisms.

The discussion on stateful reputation systems highlights that reputation can derive from multiple sources, including official documents and personal traits. Ian notes the advancements in ZK proofs and their programmability, driven by recent optimizations and new ideas in the field. He discusses the development of anonymous block listing systems, which allow for efficient user banning without extensive searches.

The conversation also addresses the application of these concepts in academic contexts, where anonymity can facilitate dissent but raises concerns about bullying and trolling. A proposed solution involves self-organizing groups that manage participation democratically without knowing individual identities. The speakers discuss the utility of pseudonyms in online interactions, referencing historical examples.

Ian Miers expresses a desire to collaborate on prototyping stateful zero-knowledge identity systems, proposing methods that involve modifying existing messaging apps or using proxy servers for anonymous communication. He highlights the anthropological implications of anonymity in small groups, noting both its benefits and drawbacks, including the potential for abuse.

The discussion touches on the role of AI in moderation, including the idea of using enclaves to hold moderation keys instead of individuals. Ian acknowledges the growing definition of agentic AI and its applications in moderation, while also expressing concerns about the training and behavior of AI agents. The conversation references the prisoners' dilemma and reputation systems in anonymous environments, emphasizing the importance of reputation signals.

Ian highlights the challenges of continuous authentication in systems, emphasizing the need to prevent account sharing and credential theft. He advocates for local state management, allowing users to prove their geographic location without exposing their entire IP history. The conversation delves into identity theft concerns and the evolution of anonymous credentials into practical applications, driven by regulatory changes and increasing concerns about deep fakes and AI bots.

Miers discusses the shift towards proprietary systems versus open-source solutions, noting that data has become increasingly siloed. He emphasizes the importance of developing zk stateful software architectures that prioritize privacy without relying on specific hardware. The complexities of moderation are acknowledged, highlighting the need to balance control with appropriate content management.

The conversation centers on zk credentials and their role in identity verification, discussing Google's Longfellow and the challenge of creating unlinkable pseudonyms. Ian introduces zk promises, which allow for mutable identities based on reputation, and zk kill keys to monitor credential usage. He mentions an upcoming paper on private compute tokens that aims to keep reputation scores confidential.

In addressing spam detection, Ian proposes an anonymous reputation scoring system and explores the intersection of AI and bot detection. He discusses identity and reputation in the context of money laundering prevention, referencing a co-authored paper on identity composition and reputation scoring. Ian expresses enthusiasm for future experiments on anonymity and trust within small groups, encouraging listeners to engage with the concepts discussed.