When Agents are Powerful: Black Hole Search in Time-Varying Graphs

2025-10-25

AI Summary

This paper addresses the efficient localization of “black-hole nodes”—malicious nodes that consume visiting resources—in dynamic graphs. Existing approaches, constrained by strict face-to-face communication, require an excessive number of agents. To overcome this limitation, we propose an enhanced distributed search framework that equips agents with global communication capability and 1-hop neighborhood visibility, thereby relaxing strict local interaction constraints. Methodologically, our approach integrates dynamic graph traversal strategies, a local-observation-based information fusion mechanism, and a lightweight distributed coordination algorithm. Experiments demonstrate that our framework reduces agent requirements by approximately 40–60% over baselines, significantly improves task completion rate and agent survival rate, and achieves high-accuracy identification of black-hole edges. Our key contribution is the first incorporation of global communication synergized with limited local visibility into dynamic black-hole search—striking a balance among efficiency, robustness, and scalability.

Technology Category

• Search and Optimization: Distributed Search
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Multiagent Systems: Distributed Problem Solving

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Agentic search
• Social Networks and Social Media: Social mining and social search on the Web

Abstract

A black hole is a harmful node in a graph that destroys any resource entering it, making its identification a critical task. In the emph{Black Hole Search (BHS)} problem, a team of agents operates on a graph $G$ with the objective that at least one agent must survive and correctly identify an edge incident to the black hole. Prior work has addressed BHS in arbitrary dynamic graphs under the restrictive emph{face-to-face} communication, where agents can exchange information only when co-located. This constraint significantly increases the number of agents required to solve the problem. In this work, we strengthen the capabilities of agents in two ways: (i) granting them emph{global communication}, enabling interaction regardless of location, and (ii) equipping them with emph{1-hop visibility}, allowing each agent to observe its immediate neighborhood. These enhancements lead to more efficient solutions for the BHS problem in dynamic graphs.

Problem

• Identifying harmful nodes in dynamic graphs using enhanced agents
• Overcoming face-to-face communication limits with global interaction
• Reducing agent requirements through 1-hop neighborhood visibility

Innovation

• Agents use global communication for remote interaction
• Agents employ 1-hop visibility to observe neighborhoods
• Enhanced capabilities solve Black Hole Search efficiently