CILC Cryptographically-secure Inter-agent Loop Closure Candidate Detection for Multi-Agent Collaborative SLAM

Jul 16, 2026

Upload by Skyler

Multi-agent SLAM and collaborative SLAM depend on sharing global descriptors to detect inter-agent loop closures, but conventional encrypted radios only protect against external eavesdropping and do not stop a compromised robot from exploiting public broadcasts. CILC (Cryptographically-secure Inter-agent Loop Closure candidate detection) uses secure multi-party computation to perform descriptor similarity checks without disclosing descriptors in the clear. The system secures only the candidate-detection stage for an efficient privacy–overhead trade-off, while validation shows real-time, communication-feasible operation across visual and LiDAR descriptors.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages9

Reading Time23 min

Abstract

Introduction

CILC: Cryptographically-secure Inter-agent Loop Closure Candidate Detection for Multi-Agent Collaborative SLAM Andrew Fishberg 1 , Yixuan Jia 1 , Jonathan P. How 1 arXiv :2607 .06700v 1 [ cs .RO] 7 Jul 2026 Abstract—Multi-agent Simultaneous Localization and Mapping (SLAM) and collaborative SLAM (CSLAM) require robots to continuously exchange global descriptors (GDs) to detect inter-agent loop closures (ILCs) . While encrypted radios protect this traffic from external eavesdroppers, they offer no protection against a compromised swarm member. We show this threat is concrete by demonstrating how a corrupted agent can reconstruct approximations of an honest agent’s imagery and trajectory from its public GD broadcasts. To address this, we propose CILC (Cryptographically-secure Inter-agent Loop Closure candidate detection), a first-of-its-kind system leveraging Secure Multi-Party Computation (SMPC) to detect ILC candidates without exchanging GDs in the clear. Rather than securing the entire CSLAM pipeline, we apply SMPC only to ILC candidate detection (i.e., GD similarity comparison), a privacy-sensitive yet computationally lightweight step, yielding an advantageous privacy-to-overhead trade-off. We validate in both simulation and hardware experiments that CILC remains real-time and communication-feasible across multimodal GDs (visual and LiDAR), while mitigating information leakage to a compromised swarm agent. I. Introduction Multi-agent robotic systems offer significant advantages over single-agent systems, including distributed operation, greater spatial coverage, and improved robustness to individual failures [1] . These benefits, however, come at the cost of increased communication and information sharing among agents, creating new privacy and security challenges as larger teams collect, exchange, and coordinate using potentially sensitive data. Recent robotfleet takeovers and software supply-chain attacks make compromised agents within the swarm a realistic concern for deployed robotic systems [2], [3] . While virtually all modern multi-agent robotic systems rely on encrypted inter-agent communication, this baseline security model exposes a critical gap in the prevailing threat model. As shown in Figure 1, encryption protects data in transit from external adversaries but provides no protection during computation among coordinating agents. Standard practice therefore focuses primarily on external attacks such as eavesdropping, while remaining vulnerable to compromised agents operating within the * This work is supported by ARL DCIST under Cooperative Agreement Number W911NF-17-2-0181 . 1 Department of Aeronautics and Astronautics, Massachusetts Institute of Technology, Cambridge, MA 02139 USA.{fishberg,yixuany, [jhow}@mit.edu](jhow}@mit.edu) Fig. 1: Threat model addressed by CILC. A heterogeneous robot team performs collaborative SLAM, exchanging observations to detect inter-agent loop closures (ILCs) . Blue agents are honest collaborators, while the red agent is a compromised but otherwise legitimate swarm member. Transport-layer encryption prevents external eavesdropping, but the compromised agent can access all broadcast data and exfiltrate it to an external adversary without being detected. CILC limits information disclosure by revealing data only when sufficient evidence exists to support an ILC match. swarm. Because robots routinely exchange sensitive information, including images, locations, maps, and trajectories, without treating it as private between agents, a single compromised agent could gain access to the collective data of the entire team. Such an adversary may leak mission-critical information, disrupt coordination, manipulate shared state estimates, or otherwise act maliciously from within the system. This standard security practice in multi-agent robotics creates a inherent tension between effective collaboration and privacy: collaboration often requires robots to share information, yet

FAQs

What evidence supports CILC’s practicality for real robot teams?

The document states that CILC is validated in both simulation and hardware experiments, remaining real-time and communication-feasible across multimodal global descriptors such as visual and LiDAR.

Jul 16, 2026

Upload by Skyler