Randomness, Age, Work: Ingredients for Secure Distributed Hash Tables

Distributed Hash Tables (DHTs) are a popular and natural choice when dealing with dynamic resource location and routing. DHTs basically provide two main functions: saving (key, value) records in a network environment and, given a key, find the node responsible for it, optionally retrieving the associated value. However, all predominant DHT designs suffer a number of security flaws that expose nodes and stored data to a number of malicious attacks, ranging from disrupting correct DHT routing to corrupting data or making it unavailable. Thus even if DHTs are a standard layer for some mainstream systems (like BitTorrent or KAD clients), said vulnerabilities may prevent more security-aware systems from taking advantage of the ease of indexing and publishing on DHTs. Through the years a variety of solutions to the security flaws of DHTs have been proposed both from academia and practitioners, ranging from authentication via Central Authorities to social-network based ones. These solutions are often tailored to DHT specific implementations, simply try to mitigate without eliminating hostile actions aimed at resources or nodes. Moreover all these solutions often sports serious limitations or make strong assumptions on the underlying network. We present, after after providing a useful abstract model of the DHT protocol and infrastructure, two new primitives. We extend a “standard” proof-of-work primitive making of it also a “proof of age” primitive (informally, allowing a node to prove it is “sufficiently old”) and a “shared random seed” primitive (informally, producing a new, shared, seed that was completely unpredictable in a “sufficiently remote” past). These primitives are then integrated into the basic DHT model obtaining an “enhanced” DHT design, resilient to many common attacks. This work also shows how to adapt a Block Chain scheme – a continuously growing list of records (or blocks) protected from alteration or forgery – to provide a possible infrastructure for our proposed secure design. Finally a working proof-of-concept software implementing an “enhanced” Kademlia-based DHT is presented, together with some experimental results showing that, in practice, the performance overhead of the additional security layer is more than tolerable. Therefore this work provides a threefold contribution. It describes a general set of new primitives (adaptable to any DHT matching our basic model) achieving a secure DHT; it proposes an actionable design to attain said primitives; it makes public a proof-of-concept implementation of a full “enhanced” DHT system, which a preliminary performance evaluation shows to be actually usable in practice