Vincenzo Sammartino
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Vincenzo Sammartino has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
- Fabrizio Baiardi
Università di Pisa
6 shared publications - Salvatore Ruggieri
University of Pisa, Pisa, Italy
3 shared publications - Andrea Frasson
University of Pisa
2 shared publications - Jacopo Gneri
University of Pisa
2 shared publications - Valeria Picchianti
University of Pisa
2 shared publications - Abhishek Kumar Pandey
International Institute of Information Technology
1 shared publication
Latest Publications
- 7 October 2026
- Published by Springer Nature
- p. 629-649
Fight Drones with Drones: Detecting Aerial Perimeter Intrusion Using Drone-Mounted Microphones
- 7 October 2026
- Published by Springer Nature
- p. 629-649
Fight Drones with Drones: Detecting Aerial Perimeter Intrusion Using Drone-Mounted Microphones
- 25 September 2026
- Published by Elsevier in Knowledge-Based Systems
- Vol. 352
Vulnerabilities in autonomous execution: A survey of security threats and defenses in LLM-driven agentic and multi-agent systems
- 25 September 2026
- Published by Elsevier in Knowledge-Based Systems
- Vol. 352
Vulnerabilities in autonomous execution: A survey of security threats and defenses in LLM-driven agentic and multi-agent systems
- Open Access
- 31 August 2026
- Published by Elsevier in SSRN Electronic Journal
Adversarial Reliability: From Design to Deployment and Beyond Using Attack-Path Chokepoints
Most standard performance models neglect adversarial reliability, the probability that a system behaves correctly under attack. We propose a framework to assess and improve it centred on chokepoints, intersections of intrusions, where single defensive updates disproportionately constrain an adversary. Our approach relies on a three-layer ontology: a quantitative security twin of the infrastructure, Monte Carlo simulations of attacker strategies to build intrusions, and an intrusion graph that merges the intrusions to discover chokepoints. From this graph, we estimate metrics such as the traversal frequency, the Mean-Time-To-Compromise (MTTC), the probability of breach Pbreach(t) and the availability-under-attack Aadv. This common foundation unifies seven security problems into chokepoint analyses, including network segmentation, IDS and honeypot allocation, forensic reconstruction, and global joint budget optimisation against zero-day vulnerabilities. Because the security twin can be built from architectural specifications and CVE databases, our approach can assess both security-by-design and operational adversarial reliability. Experiments on the security twins of two real systems and on synthetic topologies up to |V| = 10,000 show significant reliability improvements: at a representative security budget equal to 5% of the total deployment cost of the assets of the system, the joint allocation increases the MTTC by 3.9× over the undefended baseline, and the design-time predictions match post-deployment measurements within 8%.
- 31 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessAdversarial Reliability: From Design to Deployment and Beyond Using Attack-Path Chokepoints
Most standard performance models neglect adversarial reliability, the probability that a system behaves correctly under attack. We propose a framework to assess and improve it centred on chokepoints, intersections of intrusions, where single defensive updates disproportionately constrain an adversary. Our approach relies on a three-layer ontology: a quantitative security twin of the infrastructure, Monte Carlo simulations of attacker strategies to build intrusions, and an intrusion graph that merges the intrusions to discover chokepoints. From this graph, we estimate metrics such as the traversal frequency, the Mean-Time-To-Compromise (MTTC), the probability of breach Pbreach(t) and the availability-under-attack Aadv. This common foundation unifies seven security problems into chokepoint analyses, including network segmentation, IDS and honeypot allocation, forensic reconstruction, and global joint budget optimisation against zero-day vulnerabilities. Because the security twin can be built from architectural specifications and CVE databases, our approach can assess both security-by-design and operational adversarial reliability. Experiments on the security twins of two real systems and on synthetic topologies up to |V| = 10,000 show significant reliability improvements: at a representative security budget equal to 5% of the total deployment cost of the assets of the system, the joint allocation increases the MTTC by 3.9× over the undefended baseline, and the design-time predictions match post-deployment measurements within 8%.
- Open Access
- 2 August 2026
- Published by Elsevier in SSRN Electronic Journal
Vulnerabilities in Autonomous Execution: A Survey of Security Threats and Defenses in LLM-driven Multi-Agent Systems
Large language models (LLMs) are no longer confined to producing text. Equipped with tools and composed into multi-agent systems, they now act on the world with limited human oversight. This shift from generation to autonomous execution transforms the security landscape. While prompt injection against a chatbot yields, at worst, an embarrassing sentence, the same injection against a tool-using agent can result in an exfiltrated database, a fraudulent transaction, or a poisoned message propagating to peer agents. The vulnerability class is not new, but the blast radius is.This article surveys the security of LLM-driven agentic and multi-agent systems from 2023 through early 2026. We contribute a threat taxonomy classifying attacks by entry point, propagation path, and target asset. Using this, we organize four threat families—indirect prompt injection, malicious tool use, inter-agent infection, and memory poisoning—consolidating 44 attack papers into a unified comparison. We review the defensive landscape across five categories (detection, structural separation, execution isolation, information-flow control, and alignment-side hardening) and map defenses to threats to expose coverage gaps.Two findings recur. First, no purely prompt-level defense withstands adaptive attackers; effective security requires constraining what the system around the model is permitted to do, rather than relying on model robustness. Second, multi-agent propagation—the mechanism making agentic compromise qualitatively worse than single-model compromise—is the least defended part of the stack, with almost no deployed frameworks enforcing trust boundaries between cooperating agents. We conclude with a research agenda and a minimal security-reporting standard for agent frameworks.
- 2 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessVulnerabilities in Autonomous Execution: A Survey of Security Threats and Defenses in LLM-driven Multi-Agent Systems
Large language models (LLMs) are no longer confined to producing text. Equipped with tools and composed into multi-agent systems, they now act on the world with limited human oversight. This shift from generation to autonomous execution transforms the security landscape. While prompt injection against a chatbot yields, at worst, an embarrassing sentence, the same injection against a tool-using agent can result in an exfiltrated database, a fraudulent transaction, or a poisoned message propagating to peer agents. The vulnerability class is not new, but the blast radius is.This article surveys the security of LLM-driven agentic and multi-agent systems from 2023 through early 2026. We contribute a threat taxonomy classifying attacks by entry point, propagation path, and target asset. Using this, we organize four threat families—indirect prompt injection, malicious tool use, inter-agent infection, and memory poisoning—consolidating 44 attack papers into a unified comparison. We review the defensive landscape across five categories (detection, structural separation, execution isolation, information-flow control, and alignment-side hardening) and map defenses to threats to expose coverage gaps.Two findings recur. First, no purely prompt-level defense withstands adaptive attackers; effective security requires constraining what the system around the model is permitted to do, rather than relying on model robustness. Second, multi-agent propagation—the mechanism making agentic compromise qualitatively worse than single-model compromise—is the least defended part of the stack, with almost no deployed frameworks enforcing trust boundaries between cooperating agents. We conclude with a research agenda and a minimal security-reporting standard for agent frameworks.
- 8 June 2026
- Published by Elsevier in Future Generation Computer Systems
- Vol. 185
Hybrid quantum Graph Neural Networks for robust botnet detection in modern IoT ecosystems
- 8 June 2026
- Published by Elsevier in Future Generation Computer Systems
- Vol. 185
Hybrid quantum Graph Neural Networks for robust botnet detection in modern IoT ecosystems
- Top Cited Papers
- 9 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Communications Standards Magazine
- Vol. PP (24712825) , 1-7
Blockchain-Enabled Quantum-Resistant Hybrid Isogeny-Based Signcryption Framework for Digital Twin-Assisted Healthcare Systems
The rapid proliferation of the Internet of Medical Things (IoMT) has facilitated the development of Digital Twin (DT) technology in healthcare, enabling real-time simulation and remote patient monitoring. However, the reliance on public communication channels exposes sensitive physiological data to severe security threats. Furthermore, the imminent advent of quantum computing renders traditional public-key cryptographic primitives (e.g., RSA and Elliptic Curve Cryptography (ECC)) vulnerable to Shor’s algorithm. To address these challenges, this article proposes a blockchain-enabled, quantum-resistant hybrid signcryption framework tailored for DT-assisted healthcare. We integrate Isogeny-Based Cryptography (IBC) signcryption that leverages the Commutative Supersingular Isogeny Diffie-Hellman (CSIDH) problem to ensure simultaneous confidentiality and unforgeability with post-quantum security guarantees. The proposed framework utilizes a private blockchain to ensure the immutability and traceability of medical data transactions between the physical patient and their digital twin. Simulation results demonstrate the blockchain’s scalability and show that, while isogeny-based primitives introduce computational overhead compared to classical methods, the scheme remains feasible for high-value medical data synchronization.
- 9 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Communications Standards Magazine
- Vol. PP (24712825) , 1-7
Top Cited PapersBlockchain-Enabled Quantum-Resistant Hybrid Isogeny-Based Signcryption Framework for Digital Twin-Assisted Healthcare Systems
The rapid proliferation of the Internet of Medical Things (IoMT) has facilitated the development of Digital Twin (DT) technology in healthcare, enabling real-time simulation and remote patient monitoring. However, the reliance on public communication channels exposes sensitive physiological data to severe security threats. Furthermore, the imminent advent of quantum computing renders traditional public-key cryptographic primitives (e.g., RSA and Elliptic Curve Cryptography (ECC)) vulnerable to Shor’s algorithm. To address these challenges, this article proposes a blockchain-enabled, quantum-resistant hybrid signcryption framework tailored for DT-assisted healthcare. We integrate Isogeny-Based Cryptography (IBC) signcryption that leverages the Commutative Supersingular Isogeny Diffie-Hellman (CSIDH) problem to ensure simultaneous confidentiality and unforgeability with post-quantum security guarantees. The proposed framework utilizes a private blockchain to ensure the immutability and traceability of medical data transactions between the physical patient and their digital twin. Simulation results demonstrate the blockchain’s scalability and show that, while isogeny-based primitives introduce computational overhead compared to classical methods, the scheme remains feasible for high-value medical data synchronization.
- Open Access
- 19 January 2026
- Published by Springer Nature in Research Square
Simulation-Powered Cybersecurity: Real-Time Risk Assessment via Non-Intrusive Security Twin
Digital twin technology is emerging as the cornerstone of next-generation cybersecurity. A security twin is a graph-based model which acts as a dynamic inventory enriched with vulnerability intelligence and that can mirror complex ICT infrastructures to predict the behavior of threats without disrupting live production environments. However, maintaining high-fidelity synchronisation between the infrastructure and the twin remains a challenge, mainly when active scanning cannot be employed.This paper introduces NotLine , a non-intrusive and fully automated platform that builds and updates a security twin through the continuous passive ingestion of multi-protocol network telemetry. NotLine leverages a distributed monitoring pipeline architecture to filter, normalize, and correlate heterogeneous traffic metadata in real-time. NotLine maps these data to the security twin.The core innovation of NotLine lies in its integration of this live model with an AI-driven Monte Carlo simulation engine . The engine uses the security twin to generate the state transitions of a threat actor as determined by the access rights and information it has acquired. This enables the quantification of risk exposure probabilistically and enables prescriptive analytics and preemptive remediation.We present a comprehensive evaluation of NotLine in a production environment and show that a hypoexponential mathematical model characterises the platform discovery pattern. According to this model, the platform maps the majority of assets within 48 hours, but a long-tail monitoring period is critical to capture all infrastructure components. These results confirm that NotLine provides a robust foundation for simulation-powered cybersecurity, bridging the gap between passive observation and proactive risk prediction.
- 19 January 2026
- Published by Springer Nature in Research Square
Open AccessSimulation-Powered Cybersecurity: Real-Time Risk Assessment via Non-Intrusive Security Twin
Digital twin technology is emerging as the cornerstone of next-generation cybersecurity. A security twin is a graph-based model which acts as a dynamic inventory enriched with vulnerability intelligence and that can mirror complex ICT infrastructures to predict the behavior of threats without disrupting live production environments. However, maintaining high-fidelity synchronisation between the infrastructure and the twin remains a challenge, mainly when active scanning cannot be employed.This paper introduces NotLine , a non-intrusive and fully automated platform that builds and updates a security twin through the continuous passive ingestion of multi-protocol network telemetry. NotLine leverages a distributed monitoring pipeline architecture to filter, normalize, and correlate heterogeneous traffic metadata in real-time. NotLine maps these data to the security twin.The core innovation of NotLine lies in its integration of this live model with an AI-driven Monte Carlo simulation engine . The engine uses the security twin to generate the state transitions of a threat actor as determined by the access rights and information it has acquired. This enables the quantification of risk exposure probabilistically and enables prescriptive analytics and preemptive remediation.We present a comprehensive evaluation of NotLine in a production environment and show that a hypoexponential mathematical model characterises the platform discovery pattern. According to this model, the platform maps the majority of assets within 48 hours, but a long-tail monitoring period is critical to capture all infrastructure components. These results confirm that NotLine provides a robust foundation for simulation-powered cybersecurity, bridging the gap between passive observation and proactive risk prediction.
- 1 January 2025
- Published by Research Publishing Services
- p. 643-650
A Security Twin to Defeat Intrusions in Cyber Physical Systems
- 1 January 2025
- Published by Research Publishing Services
- p. 643-650
A Security Twin to Defeat Intrusions in Cyber Physical Systems