Samu Varjonen
Binding Identifiers and Locators with DHT

Abstract:
It has been noted by the Internet researchers that the future internet
architecture needs to separate identifiers from the IP addresses. This
separation relieves IP addresses from the identifier task and leaves
them purely as locators for the host. Host Identity protocol (HIP)
separates the identifier from the locator. This separation is made
possible by introducing a new namespace that is based on public-private
key pairs. These cryptographic host identities are reduced to 128-bit
Host Identity Tags (HIT) by hashing, so that they can be used instead of
addresses in applications.  Mapping
between flat-namespace identities and current host locators is a
difficult task for traditional DNS systems. This is partly because of
the need for support of mobile nodes and slow update speed of
traditional DNS systems.
Distributed Hash Tables (DHT) are decentralized distributed systems,
that are organized in ring-like fashion. DHTs use consistent hashing to
distribute key-value pairs to the ring. Replication and routing schemes
supported by the DHTs make them self-healing and attack resilient. HIP
uses OpenDHT, a public DHT service
running on distributed infrastructure provided by PlanetLab.  OpenDHT
provides a simple Application Protocol Interface (API), consisting of
put, get and remove. HIP DHT interface defines how this API is used to
publish mappings from hostnames to HITs and from HITs to IP address(es).
We start with overview of Host Identity Protocol and OpenDHT, after
which we give introduction to the HIP DHT interface. HIP DHT interface
suffers from several challenges introduced by the open nature of the
OpenDHT. We describe these problems and present solutions on how to
solve them. These problems include limited lifetime of DHT entries,
spamming with bogus values. Solutions we present consider trust from the
initiator and from the responder side, depending on the use case. These
solutions use HIPs public-private keys to sign and validate information
and DHT as a decentralized storage for trust information.