Applying Multi-Agent LLMS to Threat Analysis and Risk Assessment in Automotive Security
Abstract
Threat Analysis and Risk Assessment (TARA) is essential for identifying threats and managing risks, but is often resource-intensive due to manual steps, expert involvement, and iterative revisions. To address this, we propose an automated TARA framework using LLM agents that generate threat scenarios, derive cybersecurity goals and requirements, and refine them through a writer-reviewer loop. The proof of concept, demonstrated on a keyless entry system, produces structured JSON artifacts, ensuring traceability and supporting compliance-driven cybersecurity engineering in automotive systems.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Writer-reviewer loops for automotive threat analysis produce richer attack trees.
“writer-reviewer loops for automotive threat analysis produce richer attack trees [91]”
From Agentic AI and Large Language Models for Autonomous IoT Cybersecurity: A Systematic Survey, Taxonomy, and Research Roadmap (2026) · cites this paper as [91]
Cites
Vehicle Security: A Survey of Security Issues and Vulnerabilities, Malware Attacks and Defenses
Elkhail, A. A., Refat, R. U. D., Habre, R., Hafeez, A., Bacha, A., & Malik, H. M. A. (2021) · IEEE Access · doi:10.1109/access.2021.3130495 · Cited by 87 · see citing works
A taxonomy of attack mechanisms in the automotive domain
Pekaric, I., Sauerwein, C., Haselwanter, S., & Felderer, M. (2021) · Computer Standards & Interfaces · doi:10.1016/j.csi.2021.103539 · Cited by 50 · see citing works
Safety analysis in the era of large language models: A case study of STPA using ChatGPT
Qi, Y., Zhao, X., Khastgir, S., & Huang, X. (2025) · Machine Learning with Applications · doi:10.1016/j.mlwa.2025.100622 · Cited by 39 · see citing works
ThreatSurf: A method for automated Threat Surface assessment in automotive cybersecurity engineering
Zelle, D., Plappert, C., Rieke, R., Scheuermann, D., & Krauß, C. (2022) · Microprocessors and Microsystems · doi:10.1016/j.micpro.2022.104461 · Cited by 38 · see citing works
In-Depth Exploration of ISO/SAE 21434 and Its Correlations with Existing Standards
Costantino, G., Vincenzi, M. D., & Matteucci, I. (2022) · IEEE Communications Standards Magazine · doi:10.1109/mcomstd.0001.2100080 · Cited by 33 · see citing works
Welcome Your New AI Teammate: On Safety Analysis by Leashing Large Language Models
Nouri, A., Cabrero‐Daniel, B., Törner, F., Sivencrona, H., & Berger, C. (2024) · IEEE/ACM International Conference on AI Engineering – Software Engineering for AI (CAIN) · doi:10.1145/3644815.3644953 · Cited by 22 · see citing works
Benyahya, M., Lenard, T., Collen, A., & Nijdam, N. A. (2023) · International Conference on Availability, Reliability and Security (ARES) · doi:10.1145/3600160.3605084 · Cited by 19 · see citing works
THREATGET: Towards Automated Attack Tree Analysis for Automotive Cybersecurity
Chlup, S., Christl, K., Schmittner, C., Shaaban, A. M., Schauer, S., & Latzenhofer, M. (2022) · Information · doi:10.3390/info14010014 · Cited by 15 · see citing works
Elevating TARA: A Maturity Model for Automotive Threat Analysis and Risk Assessment
Vielberth, M., RAAB, K. M., Glas, M., Grümer, P., & Pernul, G. (2024) · International Conference on Availability, Reliability and Security (ARES) · doi:10.1145/3664476.3670888 · Cited by 7 · see citing works
AI-Enhanced threat analysis and risk assessment for connected and autonomous vehicles
Ahmad, U., Han, M., & Mahmood, S. (2025) · Software Quality Journal · doi:10.1007/s11219-025-09723-6 · Cited by 4 · see citing works
Flagged first, then by citations
Cited by
Sorted by citations
Nageshwaran, V., & Ezekiel, S. (2026) · Electronics · doi:10.3390/electronics15122740 · Cited by 4 · see citing works
“(ii) Debate coordination: agents argue from different positions to surface false negatives, as in debate-driven phishing detection [90] and the writer-reviewer automotive threat analysis loop [91].”
- Authors
- Obradov, A., Pavković, B., Stojanović, D., & Četić, N.
- Affiliations
- University of Novi Sad (RS)
- University of Novi Sad Faculty of Technical Sciences,Departmant of Computing and Control Engineering,Novi Sad,Serbia
- Venue
- 2025 33rd Telecommunications Forum (TELFOR)
- Publisher
- IEEE
- Published
- 2025-11-25
- Pages
- 1-4
- OpenAlex
- W7118011241
- Cited by
- 1
- References
- 10
- Citations by year
- 2026: 1
- FWCI
- 2.98
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- Telecommunications Forum (TELFOR) (publishedVersion)
- Topics
- Physical Sciences › Engineering › Safety, Risk, Reliability and Quality › Safety Systems Engineering in Autonomy
- Physical Sciences › Computer Science › Information Systems › Information and Cyber Security
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Keywords
- threat analysis and risk assessment, automotive cybersecurity, LLM agents, keyless entry systems, traceability
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
