[Paper Review] Maximum Metric Spanning Tree made Byzantine Tolerant

Swan Dubois, Toshimitsu Masuzawa+1 more|arXiv (Cornell University)|Apr 28, 2011

Distributed systems and fault tolerance12 references4 citations

TL;DR

This paper presents the first topology-aware strongly stabilizing protocol for constructing a maximum metric spanning tree in distributed systems tolerant to an arbitrary number of Byzantine faults. It establishes necessary and sufficient conditions for such protocols to exist, proving that strong stabilization is possible if and only if the metric is strongly maximizable and the containment radius is at least max{0, |M(S)|−2}, achieving optimal fault containment with finite convergence time.

ABSTRACT

Self-stabilization is a versatile approach to fault-tolerance since it permits a distributed system to recover from any transient fault that arbitrarily corrupts the contents of all memories in the system. Byzantine tolerance is an attractive feature of distributed systems that permits to cope with arbitrary malicious behaviors. This paper focus on systems that are both self-stabilizing and Byzantine tolerant. We consider the well known problem of constructing a maximum metric tree in this context. Combining these two properties is known to induce many impossibility results. In this paper, we provide first two impossibility results about the construction of maximum metric tree in presence of transients and (permanent) Byzantine faults. Then, we provide a new self-stabilizing protocol that provides optimal containment of an arbitrary number of Byzantine faults.

Motivation & Objective

• To investigate the feasibility of combining self-stabilization and Byzantine tolerance in global distributed tasks, particularly maximum metric spanning tree construction.
• To identify fundamental impossibility results for strong stabilization in the presence of transient and Byzantine faults.
• To design a protocol that achieves optimal containment of Byzantine faults while ensuring finite convergence time.
• To establish necessary and sufficient conditions for the existence of a strongly stabilizing protocol in this context.

Proposed method

• The authors define a strongly maximizable metric as a key property enabling strong stabilization, generalizing prior results on BFS and shortest-path metrics.
• They introduce a topology-aware strong stabilization model where the containment radius depends on the network topology and metric structure.
• A novel distributed protocol is designed that uses local decision-making based on metric values and neighbor states, ensuring convergence to a maximum metric spanning tree.
• The protocol leverages a fixed-point analysis and metric monotonicity to guarantee that correct processes eventually stabilize to the correct tree structure.
• The protocol is proven to converge in finite time t = nΠ(k,δ)ΔD, where parameters depend on network diameter and metric properties.
• The solution achieves optimal containment by ensuring that Byzantine faults affect only processes within a radius of max{0, |M(S)|−2} from any correct node.

Experimental results

Research questions

• RQ1 Under what conditions is it possible to achieve strong stabilization for maximum metric spanning tree construction in the presence of Byzantine faults?
• RQ2 What is the minimal containment radius required to ensure finite disruption of correct processes due to Byzantine faults?
• RQ3 Can the containment area be made topology-aware and optimal, rather than fixed or arbitrary?
• RQ4 Is there a necessary and sufficient condition for the existence of a strongly stabilizing protocol in this setting?
• RQ5 Can the number of disruptions be bounded under weaker system specifications?

Key findings

• A necessary and sufficient condition for the existence of a (t,c,n−1)-strongly stabilizing protocol is that the metric must be strongly maximizable and the containment radius c must be at least max{0, |M(S)|−2}.
• The proposed protocol achieves optimal containment by restricting the influence of Byzantine faults to a radius of max{0, |M(S)|−2} from any correct process.
• The protocol converges in finite time t = nΠ(k,δ)ΔD, where ΔD is the diameter of the network and Π(k,δ) is a function of the metric's structure.
• The paper proves two impossibility results: strong stabilization is impossible for non-strongly maximizable metrics and for containment radii below the derived lower bound.
• Topology-aware strong stabilization is possible only when the containment area is S_B^*, the set of nodes within distance c from at least one Byzantine node, and this is optimal.
• The results close open gaps in prior work on topology-aware stabilization, particularly for general maximum metric trees.