- Rating
- 3
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
The paper investigates whether external attackers can infer a multi-robot system’s communication topology from ordinary task requests and public feedback. RoboTopo achieves promising results on the authors’ RoboCoord dataset, and the inferred graphs help attackers spread unsafe actions more effectively. The authors also propose and evaluate a mitigation method. Overall, the paper reveals an important attack surface in multi-robot collaboration, though some methodological and experimental details need further explanation to better support its conclusions.
Critical strengths
-
The paper shows that ordinary task feedback can reveal hidden communication structure through a single public interface, without access to internal messages or participant identities. This is an interesting security finding.
-
The probes are well motivated by physical constraints. Simultaneous occupancy helps estimate robot counts, while matched task controls help separate execution failures from communication limitations.
-
The attack experiments show that the inferred topology helps select more effective targets and spread unsafe behavior to more robots, demonstrating the practical security implications of topology leakage.
-
StageGate provides a useful initial defense, reducing the reconstruction score while preserving progress feedback for most requests.
Critical weaknesses
-
In the experiments verifying propagation attacks, the authors compared multiple models for InfectBot, External Entry Point Attack, and Privileged In-System Attack, whereas this paper analyzes only the performance of the GPT-4O model; the experimental results for other models are not yet known.
-
In related work, the authors described topology detection approaches similar to digital multi-agent systems (such as MASLeak and CIA), but did not clearly explain why such methods are not applicable to embodied intelligence scenarios.
-
Algorithms 1, 2, and 3 in the appendix are key manifestations of the method described in this paper; however, neither the main text nor the appendix provides an analysis of the method’s time complexity, space complexity, or probe cost. A reasonable analysis should be conducted in conjunction with actual probe scenarios.
Questions for authors
-
The entire set of topology detection and subsequent attack propagation experiments is based solely on GPT-4o. Why? Please explain how different large language models would affect the effectiveness of topology inference and attack propagation (such as InfectBot, External Entry Point Attack, and Privileged In-System Attack, as mentioned by the authors). The authors should conduct cross-model experiments to demonstrate the method’s generality;
-
For digital multi-agent topology inference methods such as MASLeak and CIA: If they can be adapted for use in this scenario, please provide comparison results showing the adapted versions; if they cannot be adapted, please explain in detail the underlying core differences and obstacles.
-
Please analyze the temporal, spatial, and probe overheads of Algorithms 1–3 in the context of real-world detection scenarios, and include an explanation of how the algorithms handle boundary cases to improve the readability of the paper.
8 comments
[1/5] Response to Reviewer 6RXY: Models and participant recovery
To W1
-
Additional target models. To align with the model coverage of the propagation studies considered in our paper, we extended topology inference evaluation to Gemini-2.5-Flash and Qwen3 across all five graph configurations, with three runs per configuration. Their mean edge F1 scores are 0.9757 and 0.9961, respectively.
-
Attack propagation. Across nine model and attack settings with GPT-4o, Gemini-2.5-Flash, and Qwen3, RoboTopo achieves the highest mean , , , and call AUC within each setting. Q1 reports all nine metrics from the main paper using the same definitions and aggregation rules.
To Q1: Topology inference
-
Model selection and additional evaluation. We selected Gemini-2.5-Flash and Qwen3 to align our topology inference evaluation with the model coverage of the propagation-attack studies considered in our paper. InfectBot includes Gemini-2.5-Flash to represent practical robotic deployments that balance latency, efficiency, and reasoning capability. The communication-attack study includes the Qwen family to reflect model backends used in commercial robot systems. We evaluated both additional models across all five graph configurations, with three runs per configuration.
-
Robot and planner counts and edge F1.
Graph GPT-4o GPT-4o GPT-4o mean edge F1 Gemini Gemini Gemini mean edge F1 Qwen Qwen Qwen mean edge F1 G0 6/0 6.3750; 6/6/6 0.2500; 0/0/0 1.0000 6.3750; 6/6/6 0.2500; 0/0/0 1.0000 6.3750; 6/6/6 0.2500; 0/0/0 1.0000 G1 7/0 7.3750; 7/7/7 0.2500; 0/0/0 0.9583 6.7917; 6/6/7 0.2500; 0/0/0 0.9387 7.3750; 7/7/7 0.2500; 0/0/0 0.9804 G2 6/1 6.3750; 6/6/6 0.7500; 1/1/1 1.0000 6.3750; 6/6/6 0.7778; 1/1/1 0.9655 6.3750; 6/6/6 0.7500; 1/1/1 1.0000 G3 4/1 4.3750; 4/4/4 0.8000; 1/1/1 1.0000 4.3750; 4/4/4 0.8000; 1/1/1 1.0000 4.3750; 4/4/4 0.8000; 1/1/1 1.0000 G4 5/2 5.3750; 5/5/5 1.5000; 2/2/2 1.0000 5.3750; 5/5/5 1.4611; 2/2/2 0.9744 5.3750; 5/5/5 1.5000; 2/2/2 1.0000 5.6000/0.8000 5.9750; 5.6000 0.7100; 0.8000 0.9917 5.8583; 5.4667 0.7078; 0.8000 0.9757 5.9750; 5.6000 0.7100; 0.8000 0.9961 -
Participant recovery and edge F1. Across the five graph configurations with three runs each, GPT-4o, Gemini-2.5-Flash, and Qwen3 recover the correct robot counts in 100% (15/15), 86.67% (13/15), and 100% (15/15) of runs, respectively. All three models recover the correct planner counts in 100% (15/15) of runs. Their mean edge F1 scores at a threshold of 0.5 are 0.9917, 0.9757, and 0.9961, respectively. Gemini’s robot-count errors are confined to G1, where it identifies six of seven robots in two runs; the third run recovers all seven robots and all nine true connections.
-
[2/5] Response to Reviewer 6RXY: Edge probabilities and coverage
To Q1 (continued): Topology inference
-
Topology inference (continued).
-
Edge probabilities and coverage (Table 2 metrics).
Graph Edge GPT-4o GPT-4o GPT-4o Gemini Gemini Gemini Qwen Qwen Qwen G0 R→R 0.9789 0.9789 90/90 0.9524 0.9524 90/90 0.9789 0.9789 90/90 G1 R→R 0.9458 0.9458 25/27 0.9603 0.9603 24/27 0.9821 0.9821 26/27 G2 R→R 0.9848 0.9848 30/30 0.9795 0.9795 27/30 0.9861 0.9861 30/30 G2 R→P 0.7222 0.5417 9/9 0.7222 0.5617 9/9 0.7222 0.5417 9/9 G2 P→R 0.7500 0.5625 9/9 0.7407 0.5756 9/9 0.7500 0.5625 9/9 G3 R→P 0.7500 0.6000 12/12 0.7500 0.6000 12/12 0.7153 0.5722 12/12 G3 P→R 0.8000 0.6400 12/12 0.8000 0.6400 12/12 0.7653 0.6122 12/12 G4 R→R 0.9844 0.9844 18/18 0.9811 0.9811 18/18 0.9811 0.9811 18/18 G4 R→P 0.7083 0.5313 12/12 0.7023 0.5096 11/12 0.7083 0.5313 12/12 G4 P→R 0.7083 0.5313 12/12 0.7023 0.5096 11/12 0.7083 0.5313 12/12 R→R 0.9735 0.9735 163/165 0.9683 0.9683 159/165 0.9820 0.9820 164/165 R→P 0.7269 0.5576 33/33 0.7248 0.5571 32/33 0.7153 0.5484 33/33 P→R 0.7528 0.5779 33/33 0.7477 0.5751 32/33 0.7412 0.5687 33/33 -
Connection recovery. Gemini and Qwen achieve robot-to-robot connection coverage of 96.36% (159/165) and 99.39% (164/165), respectively, compared with 98.79% (163/165) for GPT-4o. For both robot-to-planner and planner-to-robot connections, Gemini achieves 96.97% (32/33) coverage, while Qwen and GPT-4o achieve 100% (33/33). On supported true robot-to-robot connections, the mean conditional probabilities averaged across applicable graphs are 0.9683, 0.9820, and 0.9735 for Gemini, Qwen, and GPT-4o, respectively. These results show that topology recovery remains effective with both additional models, while performance varies across graph configurations.
-
Summary of Revisions for Reviewer 6RXY
Thank you for your review and support of our work. In response to your comments, we have made the following revisions to the manuscript.