Larry Richard Carley
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Larry Richard Carley has a total of 34 co-authors. Here are the top co-authors by number of shared publications:
- Kathleen M. Carley
Carnegie Mellon University
16 shared publications - Jeongkeun Shin
Carnegie Mellon University
7 shared publications - Ghita Mezzour
International University of Rabat
2 shared publications - Jeyanandh Paramesh
Department of Electrical and Computer Engineering, Texas AM University, College Station, TX, USA
2 shared publications - Miray Kas
Carnegie Mellon University
2 shared publications - Miray Kas
Carnegie Mellon University Pittsburgh PA15213 USA
2 shared publications
Latest Publications
- 9 October 2026
- Published by Springer Nature
- p. 163-173
Network-Based Cost-Aware Optimization of Ransomware Mitigation Strategies
- 9 October 2026
- Published by Springer Nature
- p. 163-173
Network-Based Cost-Aware Optimization of Ransomware Mitigation Strategies
- 9 October 2026
- Published by Springer Nature
- p. 152-162
Network-Based Defense Optimization Against Ransomware Campaigns
- 9 October 2026
- Published by Springer Nature
- p. 152-162
Network-Based Defense Optimization Against Ransomware Campaigns
- Top Cited Papers
- 3 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network Science and Engineering
- Vol. 13, 8241-8260
Cyber Attack Flow Intelligence Network: Backbone Analysis and Defense Strategy Optimization
Cyberattacks increasingly unfold as multi-stage campaigns, creating a practical challenge for defenders with limited budgets: where to place controls to disrupt adversary pathways most effectively. To address this question, we construct a Cyber Attack Flow Intelligence Network using 1,031 technical and analytical reports mapped to 170 MITRE ATT&CK groups and 50 campaigns. Using this dataset, we first apply backbone analysis to identify the core patterns of attack progression and the cross-domain transition patterns. We then compute the minimum multi-vertex cut to determine the smallest optimal set of blocking techniques that can interrupt the paths leading from commonly observed initial-access techniques to high-impact objectives such as ransomware, data destruction, and financial theft. We also interpret the cost of blocking a specific technique as a defensive cost and formulate defense planning as a budget-constrained interdiction problem. Our analysis shows that when security budgets are sufficient, the most effective strategy is to build technically robust defenses that disrupt post-intrusion activity. When budgets are limited, reducing human-driven vulnerabilities through employee security training and preventing initial malware execution becomes the most effective approach. These findings provide a data-driven basis for prioritizing security investments.
- 3 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network Science and Engineering
- Vol. 13, 8241-8260
Top Cited PapersCyber Attack Flow Intelligence Network: Backbone Analysis and Defense Strategy Optimization
Cyberattacks increasingly unfold as multi-stage campaigns, creating a practical challenge for defenders with limited budgets: where to place controls to disrupt adversary pathways most effectively. To address this question, we construct a Cyber Attack Flow Intelligence Network using 1,031 technical and analytical reports mapped to 170 MITRE ATT&CK groups and 50 campaigns. Using this dataset, we first apply backbone analysis to identify the core patterns of attack progression and the cross-domain transition patterns. We then compute the minimum multi-vertex cut to determine the smallest optimal set of blocking techniques that can interrupt the paths leading from commonly observed initial-access techniques to high-impact objectives such as ransomware, data destruction, and financial theft. We also interpret the cost of blocking a specific technique as a defensive cost and formulate defense planning as a budget-constrained interdiction problem. Our analysis shows that when security budgets are sufficient, the most effective strategy is to build technically robust defenses that disrupt post-intrusion activity. When budgets are limited, reducing human-driven vulnerabilities through employee security training and preventing initial malware execution becomes the most effective approach. These findings provide a data-driven basis for prioritizing security investments.
- 8 October 2025
- Published by Springer Nature
- p. 66-75
Network Analysis of Attack Flows in Ransomware Groups and Campaigns
- 8 October 2025
- Published by Springer Nature
- p. 66-75
Network Analysis of Attack Flows in Ransomware Groups and Campaigns
- Open Access
- 24 March 2025
- Published by Academic Conferences and Publishing International Ltd in International Conference on Cyber Warfare and Security
- Vol. 20 (1) , 614-623
Simulation of Human Organizations with Computational Human Factors Against Phishing Campaigns
Traditionally, cybersecurity has focused on identifying and addressing system-level vulnerabilities that cybercriminals could exploit. As technical defenses have become more sophisticated, cybercriminals have shifted their tactics toward exploiting human users through social engineering techniques. This shift demonstrates how a single mistake by an individual within an organization can allow attackers to bypass even the most robust cybersecurity systems. Consequently, researchers have long sought to understand which human factors make individuals more susceptible to social engineering attacks. While the relationship between susceptibility to social engineering attacks and static human factors, such as age, gender, and personality, has been widely explored in empirical studies, research into the relationship between dynamic human factors, such as fatigue, perceived vulnerability, and job performance, and susceptibility to social engineering tactics has been limited. To address this gap, we propose a simulation-based methodology to explore how dynamic human factors correlate with susceptibility to spearphishing, one of the most prevalent forms of social engineering. In this study, we replicate a real-world human organization that was previously the subject of a spearphishing empirical study. Then, we computationally model dynamic human factors such as fatigue, perceived vulnerability, and job performance by integrating regression models from various human factors studies. Next, we simulate spearphishing attacks using different combinations of dynamic human factor values to explore their relationship with susceptibility to these attacks. Our simulation study reveals that when end users within an organization exhibit higher perceived vulnerability, higher job performance, and lower fatigue, they are more likely to adhere to security policies, which in turn results in both the overall number of users tricked by a spearphishing campaign and the total amount of exfiltrated data decreasing. Based on these hypotheses derived from simulation results and statistical analysis, we recommend which organizational policies should be prioritized to effectively mitigate spearphishing risks.
- 24 March 2025
- Published by Academic Conferences and Publishing International Ltd in International Conference on Cyber Warfare and Security
- Vol. 20 (1) , 614-623
Open AccessSimulation of Human Organizations with Computational Human Factors Against Phishing Campaigns
Traditionally, cybersecurity has focused on identifying and addressing system-level vulnerabilities that cybercriminals could exploit. As technical defenses have become more sophisticated, cybercriminals have shifted their tactics toward exploiting human users through social engineering techniques. This shift demonstrates how a single mistake by an individual within an organization can allow attackers to bypass even the most robust cybersecurity systems. Consequently, researchers have long sought to understand which human factors make individuals more susceptible to social engineering attacks. While the relationship between susceptibility to social engineering attacks and static human factors, such as age, gender, and personality, has been widely explored in empirical studies, research into the relationship between dynamic human factors, such as fatigue, perceived vulnerability, and job performance, and susceptibility to social engineering tactics has been limited. To address this gap, we propose a simulation-based methodology to explore how dynamic human factors correlate with susceptibility to spearphishing, one of the most prevalent forms of social engineering. In this study, we replicate a real-world human organization that was previously the subject of a spearphishing empirical study. Then, we computationally model dynamic human factors such as fatigue, perceived vulnerability, and job performance by integrating regression models from various human factors studies. Next, we simulate spearphishing attacks using different combinations of dynamic human factor values to explore their relationship with susceptibility to these attacks. Our simulation study reveals that when end users within an organization exhibit higher perceived vulnerability, higher job performance, and lower fatigue, they are more likely to adhere to security policies, which in turn results in both the overall number of users tricked by a spearphishing campaign and the total amount of exfiltrated data decreasing. Based on these hypotheses derived from simulation results and statistical analysis, we recommend which organizational policies should be prioritized to effectively mitigate spearphishing risks.
- 20 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-13
Simulation-Based Study on False Alarms in Intrusion Detection Systems for Organizations Facing Dual Phishing and Dos Attacks
Machine learning-based intrusion detection systems (IDS) have attracted considerable attention for their role in proactively identifying intrusion attempts and facilitating swift organizational response. While numerous researchers have conceptually discussed the potential negative impacts of high false alarms in machine learning-based IDS on organizations and have proposed various methods to reduce them, there is a shortage of studies that explore how these false alarms can exacerbate cyberattack damage in different organizational settings and in the face of various cyberattack campaigns. This paper introduces an agent-based modeling and simulation approach to assess false alarm consequences in machine learning-based IDS during dual Denial of Service (DoS) and phishing attacks. The IDS with distinct false positive rates, constructed using the KDD Cup 1999 dataset with diverse machine learning algorithms, were simulated to analyze how these varying false alarm rates affect the extent of damage caused by phishing and DoS attacks.
- 20 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-13
Simulation-Based Study on False Alarms in Intrusion Detection Systems for Organizations Facing Dual Phishing and Dos Attacks
Machine learning-based intrusion detection systems (IDS) have attracted considerable attention for their role in proactively identifying intrusion attempts and facilitating swift organizational response. While numerous researchers have conceptually discussed the potential negative impacts of high false alarms in machine learning-based IDS on organizations and have proposed various methods to reduce them, there is a shortage of studies that explore how these false alarms can exacerbate cyberattack damage in different organizational settings and in the face of various cyberattack campaigns. This paper introduces an agent-based modeling and simulation approach to assess false alarm consequences in machine learning-based IDS during dual Denial of Service (DoS) and phishing attacks. The IDS with distinct false positive rates, constructed using the KDD Cup 1999 dataset with diverse machine learning algorithms, were simulated to analyze how these varying false alarm rates affect the extent of damage caused by phishing and DoS attacks.
- 29 February 2024
- Published by Springer Nature
- p. 328-339
Improved Change Detection in Longitudinal Social Network Measures Subject to Pattern-of-Life Variations
- 29 February 2024
- Published by Springer Nature
- p. 328-339
Improved Change Detection in Longitudinal Social Network Measures Subject to Pattern-of-Life Variations
- 16 September 2023
- Published by Springer Nature
- p. 169-178
Integrating Human Factors into Agent-Based Simulation for Dynamic Phishing Susceptibility
- 16 September 2023
- Published by Springer Nature
- p. 169-178
Integrating Human Factors into Agent-Based Simulation for Dynamic Phishing Susceptibility
- 18 September 2022
- Published by Springer Nature
- p. 134-143
OSIRIS: Organization Simulation in Response to Intrusion Strategies
- 18 September 2022
- Published by Springer Nature
- p. 134-143
OSIRIS: Organization Simulation in Response to Intrusion Strategies
- 15 June 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems II: Express Briefs
- Vol. 69 (9) , 3665-3669
A Deep Reinforcement Learning Framework for High-Dimensional Circuit Linearization
Despite the successes of Reinforcement Learning (RL) in recent years, tasks that require exploring over long trajectories with limited feedback and searching in high-dimensional space remain challenging. This brief proposes a deep RL framework for high-dimensional circuit linearization with an efficient exploration strategy leveraging a scaled dot-product attention scheme and search on the replay technique. As a proof of concept, a 5-bit digital-to-time converter (DTC) is built as the environment, and an RL agent learns to tune the calibration words of the delay stages to minimize the integral nonlinearity (INL) with only scalar feedback. The policy network which selects calibration words is trained by the Soft Actor-Critic (SAC) algorithm. Our results show that the proposed RL framework can reduce the INL to less than 0.5 LSB within 60, 000 trials, which is much smaller than the size of searching space.
- 15 June 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems II: Express Briefs
- Vol. 69 (9) , 3665-3669
A Deep Reinforcement Learning Framework for High-Dimensional Circuit Linearization
Despite the successes of Reinforcement Learning (RL) in recent years, tasks that require exploring over long trajectories with limited feedback and searching in high-dimensional space remain challenging. This brief proposes a deep RL framework for high-dimensional circuit linearization with an efficient exploration strategy leveraging a scaled dot-product attention scheme and search on the replay technique. As a proof of concept, a 5-bit digital-to-time converter (DTC) is built as the environment, and an RL agent learns to tune the calibration words of the delay stages to minimize the integral nonlinearity (INL) with only scalar feedback. The policy network which selects calibration words is trained by the Soft Actor-Critic (SAC) algorithm. Our results show that the proposed RL framework can reduce the INL to less than 0.5 LSB within 60, 000 trials, which is much smaller than the size of searching space.