ELASTICITY: Topological Characterization of Robustness in Complex Networks

Just as a herd of animals relies on its robust social structure to survive in the wild, similarly robustness is a crucial characteristic for the survival of a complex network under attack. The capacity to measure robustness in complex networks defines the resolve of a network to maintain functionality in the advent of classical component failures and at the onset of cryptic malicious attacks. To date, robustness metrics are deficient and unfortunately the following dilemmas exist: accurate models necessitate complex analysis while conversely, simple models lack applicability to our definition of robustness. In this paper, we define robustness and present a novel metric, elasticity- a bridge between accuracy and complexity-a link in the chain of network robustness. Additionally, we explore the performance of elasticity on Internet topologies and online social networks, and articulate results.

💡 Research Summary

The paper addresses the problem of measuring robustness in complex networks, a property that determines a network’s ability to continue delivering service under component failures or malicious attacks. Existing robustness metrics either require heavy computational effort (e.g., various connectivity‑based measures, spectral indices) or ignore the flow‑based definition of robustness. To bridge this gap, the authors introduce a new metric called Elasticity (E), defined as the area under the curve of normalized throughput (Tp) versus the percentage of remaining nodes after successive removals. Elasticity ranges from 0 to 1; a higher value indicates that the network maintains a larger fraction of its original traffic even as many nodes or links fail.

The authors formalize the metric mathematically and propose a simple algorithm that operates solely on the network’s topology:

1. Build a shortest‑path routing matrix R(g) for the original graph.
2. Determine the bottleneck link’s maximum flow (f_max) and compute the per‑OD‑pair flow limit δ = 1/f_max.
3. For each removal step (random or targeted), recompute Tp(g) as the maximum feasible total flow under the capacity constraint.
4. Integrate Tp(g) over the fraction of surviving nodes to obtain Elasticity.

Because the method uses only adjacency information, it avoids the need for detailed router or link specifications, making it applicable even when ISP data are unavailable.

To validate Elasticity, the authors conduct extensive experiments on a variety of real‑world and synthetic topologies:

• Internet‑related graphs: HOT (Heuristically Optimal Topology), Scale‑free, BGP, Skitter, Whois, Abilene, and the synthetic Inet topology.
• Social network: MySpace, used to explore the idea of modeling the Internet with online social graphs.
• All graphs are rescaled to roughly 1,000 nodes using the d‑K series (specifically 2‑K) to preserve degree distributions and joint degree properties while keeping computational demands manageable.

Two attack scenarios are examined:

• Random failures – nodes are removed uniformly at random.
• Targeted attacks – nodes with the highest degree (or highest betweenness) are removed first.

The removal process is continued up to 80 % of the nodes, far beyond the 20 % limit used in many prior studies, thereby providing a worst‑case perspective that includes the behavior of disconnected networks.

Key findings include:

1. HOT topology exhibits the highest Elasticity among all tested networks. Its mesh‑like core provides many alternative paths, so even aggressive targeted attacks cause only a gradual decline in throughput.
2. Scale‑free networks perform poorly under targeted attacks because they rely heavily on a few high‑degree hubs; once those hubs are removed, throughput collapses rapidly, yielding low Elasticity.
3. Internet measurement graphs (BGP, Skitter, Whois, Abilene) show intermediate Elasticity values. Their performance correlates positively with the assortativity coefficient (r). Networks with higher assortativity (i.e., similar‑degree nodes tend to connect) tend to retain more flow under attack.
4. Social network (MySpace) displays relatively high Elasticity, suggesting that social graphs, which often have high clustering and assortativity, could serve as useful proxies for designing resilient communication infrastructures.
5. Comparison with prior work (e.g., the HOT vs. Scale‑free results in