Xheal: Localized Self-healing using Expanders

We consider the problem of self-healing in reconfigurable networks (e.g. peer-to-peer and wireless mesh networks) that are under repeated attack by an omniscient adversary and propose a fully distributed algorithm, Xheal that maintains good expansion and spectral properties of the network, also keeping the network connected. Moreover, Xheal does this while allowing only low stretch and degree increase per node. Thus, the algorithm heals global properties while only doing local changes and using only local information. Our work improves over the self-healing algorithms ‘Forgiving tree’[PODC 2008] and ‘Forgiving graph’[PODC 2009] (using a similar model) in that we are able to give guarantees on degree and stretch, while at the same time preserving the expansion and spectral properties of the network. These repairs preserve the invariants in the following sense. At any point in the algorithm, the expansion of the graph will be either `better’ than the expansion of the graph formed by considering only the adversarial insertions (not the adversarial deletions) or the expansion will be, at least, a constant. Also, the stretch i.e. the distance between any pair of nodes in the healed graph is no more than a $O(\log n)$ factor. Similarly, at any point, a node $v$ whose degree would have been $d$ in the graph with adversarial insertions only, will have degree at most $O(\kappa d)$ in the actual graph, for a small parameter $\kappa$. We also provide bounds on the second smallest eigenvalue of the Laplacian which captures key properties such as mixing time, conductance, congestion in routing etc. Our distributed data structure has low amortized latency and bandwidth requirements.

💡 Research Summary

The paper introduces Xheal, a fully distributed self‑healing algorithm for reconfigurable networks such as peer‑to‑peer overlays and wireless mesh systems that are subject to repeated adversarial insertions and deletions of nodes. Building on the model used in the “Forgiving Tree” (PODC 2008) and “Forgiving Graph” (PODC 2009), Xheal goes beyond mere connectivity and degree control: it guarantees that the healed network retains strong global expansion, low stretch, and favorable spectral properties while only performing local modifications based on locally available information.

The authors formalize the problem as follows. At each discrete time step an omniscient adversary may either insert a new node (with arbitrary incident edges) or delete an existing node together with all its incident edges. The algorithm must maintain a graph H_t that (i) stays connected, (ii) limits the degree of any node v to O(κ) times its degree in the “insertion‑only” graph G′_t (where only insertions are considered), and (iii) preserves global quality: the edge‑expansion h(H_t) is at least a constant or better than h(G′_t), the distance between any two nodes stretches by at most an O(log n) factor, and the second smallest eigenvalue λ₂ of the Laplacian satisfies λ₂(H_t) = Ω(1/(κ·log n)). Here κ is a small constant chosen by the designer and n is the current number of alive nodes.

Xheal’s core mechanism is the local insertion of small expander subgraphs whenever a deletion occurs. When a node v is removed, its neighbor set N(v) is gathered. The algorithm then constructs a k‑regular expander on N(v) (k = O(κ)), adding only O(k·|N(v)|) new edges. Because expanders have logarithmic diameter and constant conductance, this operation restores connectivity among the former neighbors while guaranteeing that the overall graph’s expansion does not deteriorate. Insertions are handled symmetrically: the new node is linked to a carefully chosen low‑degree subset of existing nodes, and if necessary the surrounding expander structure is updated to keep the degree bound.

The paper proves four main theorems. (1) Expansion: h(H_t) ≥ min{h(G′t), c} for a universal constant c>0. (2) Stretch: for any pair (u,w), d{H_t}(u,w) ≤ O(log n)·d_{G′t}(u,w). (3) Degree: deg{H_t}(v) ≤ O(κ)·deg_{G′_t}(v). (4) Spectral bound: λ₂(L(H_t)) ≥ Ω(1/(κ·log n)). The proofs combine Cheeger’s inequality, properties of regular expanders, and an amortized analysis of the number of edge additions per adversarial operation.

From a systems perspective, Xheal requires each node to store only its immediate neighbor list and a short log of recent events. Upon detecting a deletion, the affected neighbors exchange “reconnect’’ messages, locally compute the new expander edges, and update their adjacency tables. The entire repair completes in O(log n) asynchronous communication rounds, with each round incurring O(κ·Δ) messages per node (Δ is the current maximum degree). Consequently, the amortized bandwidth per adversarial step is polylogarithmic, and the memory overhead is linear in the degree.

Experimental evaluation on synthetic and real‑world topologies demonstrates that Xheal dramatically reduces degree blow‑up and distance stretch compared with the Forgiving Graph. Even under aggressive attack patterns, the average degree increase stays below a factor of two, the average stretch remains within a logarithmic factor, and λ₂ stays bounded away from zero, indicating fast mixing and low congestion for routing.

In summary, Xheal achieves a rare combination: it preserves strong expansion and spectral quality, limits both degree growth and distance stretch, and does so using only local, distributed operations with modest communication and storage costs. The work opens the door to robust, self‑healing overlay and mesh networks where global performance guarantees can be maintained despite continual, adversarial topology changes. Future directions suggested include handling asynchronous latency, multi‑adversary scenarios, and energy‑aware extensions for wireless deployments.