Securing Autonomous Multi-Agent AI Systems: A Zero-Trust Framework for Threat Detection, Behavioral Monitoring, and Resilience in Enterprise Environments
Nov 2026· International Journal of Scientific Research in Science and Technology· 0 citations· 8 references
Access Control and Trust
Abstract
This is an important issue in developing multi-agent systems in the context of distributed, autonomous, large-language models. Large-language model (LLM) agents delegate tasks, call tools and act on real-world agents. Generally, the most dangerous inputs arrived as natural-language messages from trusted peers, and one compromised agent could pass along its nefarious instructions to others. This paper proposes ZT-MAS, which brings together cryptographic agent identity and attestation, intent-bound capability tokens on every request, provenance tagging for taint propagation, a behavioural trust score that combines policy, sequence, provenance and drifting with taint propagation, a graduated response layer with step-up approval, restriction, quarantine and rollback, and a graduate response layer with step-down approval, restriction, quarantine and rollback. The framework is based on NIST SP 800-207, OWASP Top 10 for Agentic Applications, and MAESTRO threat model. We simulate an eight-agent enterprise workflow and sampled simulated a simulated a seeded simulation of a small company to determine the level of trust associated with goal hijacking, confused-deputy privilege escalation, prompt infection, memory poisoning, memory poisoning, and data exfiltration (500 episodes per cell). The perimeter zero-trust failed in all but behavioural tracking alone stopped every episode of a confused-deputy escalation, and behavioural monitoring stopped only volume-based exfiltration. While the perimeter and static zero-trust baselines succeeded in 99.6–100% of the episodes, ZT-MAS reduced this to 0–10.2% and blocked or held 2.1% of benign high-impact actions for approval. Many of the failures were injections manipulated to reflect the user’s real plan, and infection spread did not stop even when the user blocked the attack.
Supporting data, adapters, predictions and code for the article *Low-Cost LoRA Fine-Tuning of Small Language Models for Multi-Step Arithmetic Reasoning* by Jake O'Grady, Asena Isik Gürhan, Chee Fong Ting and Effirul Ramlan (University of Galway). We generated 20,000 GSM8K-derived arithmetic problems with step-by-step s...
O'Grady, Jake, Gürhan, Asena Isik, Chee, Fong Ting et al.· Zenodo (CERN European Organi...· 465 citations
The results are packaged in the Greenfield Startup Model (GSM), which explains the priority of startups to release the product as quickly as possible, and the need to shorten time-to-market, by speeding up the development through low-precision engineering activities.
Carmine Giardino, Nicolò Paternoster, M. Unterkalmsteiner et al.· IEEE Transactions on Softwar...· 178 citations· ⚡14
Software startup companies develop innovative, software-intensive products within limited timeframes and with few resources, searching for sustainable and scalable business models.
M. Unterkalmsteiner, P. Abrahamsson, Xiaofeng Wang et al.· e-Informatica Software Engin...· 157 citations· ⚡17
This study conducts a case survey study based on the secondary data of the major pivots happened in 49 software startups, and demonstrates that customer need pivot is the most common among all pivot types.
Sohaib Shahid Bajwa, Xiaofeng Wang, Anh Nguyen-Duc et al.· Empirical Software Engineeri...· 127 citations· ⚡15
The comparison of adopter and non-adopter sample reveals three potential adoption inhibitor, security, data privacy, and portability, which underlines the importance of the technical and security perspectives for research investigating the adoption of technology.
Nattakarn Phaphoom, Xiaofeng Wang, S. Samuel et al.· Journal of Systems and Softw...· 111 citations· ⚡8
Objective There is dearth of information on the timeliness of antenatal care (ANC) uptake. This study aimed to determine the timely ANC uptake by a medically trained provider (MTP) as per the World Health Organization (WHO) recommendations and the country guideline. Methods Cross-sectional survey was done with 2,731 wo...
B. Sarker, Musfikur Rahman, T. Rahman et al.· PLoS ONE· 89 citations· ⚡7
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduNov 20, 2026
Senior Research Scientist Jim Glass discusses how his MIT-IBM Computing Research Lab collaboration connects perception and language, and addresses the shortcoming of current models.
MIT News · Artificial Intelligence· news.mit.eduNov 20, 2026
Enforcing physics-based constraints prevents AI models from generating unreliable predictions for complex nonlinear systems in weather, climate, and turbulent flows.
An MIT and Dana-Farber Institute-led team used large language models to examine nearly 4,000 studies and hundreds of code repositories, revealing gaps in the information needed to scrutinize and reuse published research.
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.