notes: research/papers/papers-keridemlia-a-discovery-mechanism-for-keri

This data as json

path title type status classification words body
research/papers/papers-keridemlia-a-discovery-mechanism-for-keri Keridemlia: A Discovery Mechanism for KERI summary v0.1 — 2026-10-15: drafted by Claude from the PDF in the Papers repository (page images); unread by Mark; marked partner on Mark's word 2026-10-21, paraphrase only partner 580 # Keridemlia: A Discovery Mechanism for KERI **Abstract.** A short paper by Ryan W. West and Samuel M. Smith describing Keridemlia, a prototype discovery service for Key Event Receipt Infrastructure (KERI) that maps controller identifiers to witness identifiers and witness identifiers to IP addresses using a modified Kademlia distributed hash table. It covers the design, the API, verification of published data, possible optimizations and related work. For implementers interested in KERI discovery infrastructure. ## Summary The background section recounts the internet's lack of built-in security and the bolt-on fixes that followed (HTTPS with certificate authorities, DNSSEC), then introduces KERI as an end-to-end verifiable identity system built on Autonomic Identifiers (AIDs), which are self-certifying and self-managing, with a key event log (KEL) recording rotations, delegations and revocations. KERI runs in direct mode, where parties exchange IP addresses beforehand, or indirect mode, where witnesses store key events and a verifier must find a controller's witnesses and their addresses. That creates the need for discovery. Keridemlia is described as a DNS-like service that stores two mappings: controller AID to witness AIDs, and witness AID to IP address. It wraps an API around a Python implementation of Kademlia, a distributed hash table that spreads key-value pairs across nodes and can cache copies redundantly, and is designed so the hash table could be replaced. A data-structure section explains why a distributed hash table was chosen over other ways of building a distributed system. The design proceeds in three steps: integrating KERI with Kademlia, building an HTTP API (for example, endpoints to publish and fetch an AID's key event state and a witness's IP address), and verifying that data is authentic. Only signed, verified data is stored; stored signatures are returned with each GET so clients can check results themselves, which guards against cache poisoning and limits a malicious node to dropping data or denying service. The paper states that the prototype implements the first two steps and that full verification was still being worked on. Optimizations include automatic republishing when witnesses change, combining the two lookups, local caching and support for Legitimized Human Meaningful Identifiers (LIDs). Related work compares the Handshake naming network. The conclusion presents Keridemlia as making KERI's indirect mode usable and as a decentralized replacement for DNS. ## Key terms KERI, AID, key event log (KEL), key event state (KES), witness, indirect mode, discovery, Kademlia, distributed hash table (DHT), cache poisoning, end-verifiability ## Claims and their evidence - The Keridemlia prototype implemented integration with Kademlia and the API, but not yet full verification of published data. [presenter-reported] - Storing only signed, verified mappings and returning signatures with each response prevents cache poisoning on honest servers. [informative] - Kademlia was chosen after research into several approaches to distributed systems. [presenter-reported] - Unlike Handshake, Keridemlia does not depend on a single scarce namespace on a blockchain. [informative] ## Related Key Event Receipt Infrastructure (KERI) white paper, KERI specification (witnesses and indirect mode), OOBI (Out-Of-Band Introduction) specification as a later discovery mechanism. ## Provenance Drafted by Claude on 2026-10-15 from the PDF `whitepapers/keridemlia.pdf` (10 pages). Its embedded fonts do not extract as text, so pages 1 to 3, 8 and 9 were rendered as images and read; pages 4 to 7 (data-structure research, design steps 1 and 2) were read only through their headings and partial text. The PDF has no date on its face; its metadata gives the title "WIP: Keridemlia White Paper" and a creation date of 2021-02-26. Not yet read by Mark.

Links from other tables