Selected work

Representative Papers

Target-confidence Recourse Using tSeTlin machines: TRUST

Jun 17, 2026

This work addresses a critical limitation of traditional counterfactual explanations, which focus solely on flipping prediction labels while neglecting model confidence and robustness—rendering them vulnerable to perturbations in high-stakes scenarios. To overcome this, the authors propose TRUST, a novel framework that explicitly incorporates user-specified target confidence directly into the counterfactual generation process. By leveraging the interpretable clause structure of Probabilistic Tsetlin Machines (PTMs) and integrating Bayesian optimization, TRUST jointly minimizes input perturbation cost while optimizing both prediction confidence and robustness. The method achieves this by explicitly linking confidence to the stability of rule activation. Empirical results demonstrate that TRUST consistently yields highly robust counterfactuals with low recourse costs across multiple datasets; for instance, on the Haberman dataset, it attains a confidence level of 0.92 with an L2 distance of merely 0.10.

0 citationsRead paper

Demographic Patterns in Cybersecurity Culture: Insights from a Global Organisation Supporting Safety-Critical and Critical Infrastructure Sectors

Jun 12, 2026

This study investigates how demographic factors influence cybersecurity culture (CSC) within critical infrastructure organizations. Drawing on 6,502 valid survey responses from a global sample of 21,148 employees, the research employs Kruskal-Wallis tests with Dunn’s post hoc analysis to systematically assess associations between nine CSC dimensions and variables such as employment type, age, recruitment pathway, and managerial role. For the first time in safety-critical global organizations, it reveals that full-time, internally recruited, older employees and line managers consistently score significantly higher across multiple CSC dimensions compared to part-time, externally recruited, and younger counterparts. Building on these findings, the study proposes a scalable strategy that leverages high-scoring groups as cultural carriers while targeting low-scoring groups for focused interventions, thereby offering empirical grounding and actionable pathways for differentiated enhancement of organizational cybersecurity culture.

0 citationsRead paper

SALVE: Sparse Autoencoder-Latent Vector Editing for Mechanistic Control of Neural Networks

Dec 17, 2025

Deep neural networks (DNNs) suffer from poor interpretability and lack of behavioral controllability. To address this, we propose an integrated “Discover–Verify–Control” framework: first, unsupervised extraction of semantically coherent native feature bases from intermediate layers using an ℓ₁-sparse autoencoder; second, causal verification and feature-level saliency visualization via Grad-FAM, a novel gradient-based attribution method; third, precise and persistent feature editing directly in weight space, accompanied by derivation of a critical suppression threshold α_crit for fine-grained robustness diagnosis. This closed-loop paradigm enables end-to-end, verifiable, and controllable intervention—from feature discovery to causally grounded editing—for the first time. We validate cross-architectural generalizability, causal fidelity, and robustness of the edits on ResNet-18 and ViT-B/16, demonstrating substantial improvements in DNN transparency and controllability.

0 citationsRead paper

When technology is not enough: Insights from a pilot cybersecurity culture assessment in a safety-critical industrial organisation

Aug 28, 2025

Empirical research on cybersecurity culture (CSC) in safety-critical industries remains scarce. To address this gap, this study conducts a cross-national mixed-methods assessment—comprising surveys and semi-structured interviews—within a global safety-critical organization. It systematically examines employees’ security awareness, attitudes, and behavioral practices. Findings reveal that while personnel broadly value cybersecurity, their threat identification capabilities are limited, incident reporting channels are inefficient, and organizational policies exhibit substantial misalignment with operational realities. Based on these insights, the study proposes three core mechanisms for CSC enhancement: (1) sustained executive leadership engagement; (2) context-sensitive communication strategies; and (3) policy-practice alignment. This work constitutes the first empirical validation of CSC enablers and barriers within authentic industrial settings, thereby bridging a critical research void in safety-critical domains. It delivers a theoretically grounded, operationally actionable framework to foster resilient and sustainable cybersecurity culture.

0 citationsRead paper

Gaussian Process Surrogate Models for Efficient Estimation of Structural Response Distributions and Order Statistics

Mar 03, 2025

In structural serviceability limit state (SLS) assessment under vast meteorological scenarios, efficient estimation of response distributions and order statistics—e.g., the 100th-highest response (Y_{100})—remains challenging. This paper proposes a novel Gaussian process (GP)-based surrogate modeling approach that directly treats structural response as a stochastic process. By embedding the response’s probabilistic structure into the GP, the method enables analytical generation of the full response distribution and closed-form estimation of arbitrary-order order statistics—bypassing conventional Monte Carlo resampling and costly high-fidelity simulations. Integrating uncertainty quantification with finite-sample surrogate learning, the framework achieves comparable (Y_{100}) estimation accuracy to full-physics simulation using less than 1% of its computational cost, as validated on a 25-year historical meteorological dataset. The approach significantly enhances both efficiency and scalability of structural reliability assessment under extreme weather conditions.

0 citationsRead paper

Recent publications

Latest Papers

Target-confidence Recourse Using tSeTlin machines: TRUST

Jun 17, 2026

This work addresses a critical limitation of traditional counterfactual explanations, which focus solely on flipping prediction labels while neglecting model confidence and robustness—rendering them vulnerable to perturbations in high-stakes scenarios. To overcome this, the authors propose TRUST, a novel framework that explicitly incorporates user-specified target confidence directly into the counterfactual generation process. By leveraging the interpretable clause structure of Probabilistic Tsetlin Machines (PTMs) and integrating Bayesian optimization, TRUST jointly minimizes input perturbation cost while optimizing both prediction confidence and robustness. The method achieves this by explicitly linking confidence to the stability of rule activation. Empirical results demonstrate that TRUST consistently yields highly robust counterfactuals with low recourse costs across multiple datasets; for instance, on the Haberman dataset, it attains a confidence level of 0.92 with an L2 distance of merely 0.10.

0 citationsRead paper

Demographic Patterns in Cybersecurity Culture: Insights from a Global Organisation Supporting Safety-Critical and Critical Infrastructure Sectors

Jun 12, 2026

This study investigates how demographic factors influence cybersecurity culture (CSC) within critical infrastructure organizations. Drawing on 6,502 valid survey responses from a global sample of 21,148 employees, the research employs Kruskal-Wallis tests with Dunn’s post hoc analysis to systematically assess associations between nine CSC dimensions and variables such as employment type, age, recruitment pathway, and managerial role. For the first time in safety-critical global organizations, it reveals that full-time, internally recruited, older employees and line managers consistently score significantly higher across multiple CSC dimensions compared to part-time, externally recruited, and younger counterparts. Building on these findings, the study proposes a scalable strategy that leverages high-scoring groups as cultural carriers while targeting low-scoring groups for focused interventions, thereby offering empirical grounding and actionable pathways for differentiated enhancement of organizational cybersecurity culture.

0 citationsRead paper

SALVE: Sparse Autoencoder-Latent Vector Editing for Mechanistic Control of Neural Networks

Dec 17, 2025

Deep neural networks (DNNs) suffer from poor interpretability and lack of behavioral controllability. To address this, we propose an integrated “Discover–Verify–Control” framework: first, unsupervised extraction of semantically coherent native feature bases from intermediate layers using an ℓ₁-sparse autoencoder; second, causal verification and feature-level saliency visualization via Grad-FAM, a novel gradient-based attribution method; third, precise and persistent feature editing directly in weight space, accompanied by derivation of a critical suppression threshold α_crit for fine-grained robustness diagnosis. This closed-loop paradigm enables end-to-end, verifiable, and controllable intervention—from feature discovery to causally grounded editing—for the first time. We validate cross-architectural generalizability, causal fidelity, and robustness of the edits on ResNet-18 and ViT-B/16, demonstrating substantial improvements in DNN transparency and controllability.

0 citationsRead paper

When technology is not enough: Insights from a pilot cybersecurity culture assessment in a safety-critical industrial organisation

Aug 28, 2025

Empirical research on cybersecurity culture (CSC) in safety-critical industries remains scarce. To address this gap, this study conducts a cross-national mixed-methods assessment—comprising surveys and semi-structured interviews—within a global safety-critical organization. It systematically examines employees’ security awareness, attitudes, and behavioral practices. Findings reveal that while personnel broadly value cybersecurity, their threat identification capabilities are limited, incident reporting channels are inefficient, and organizational policies exhibit substantial misalignment with operational realities. Based on these insights, the study proposes three core mechanisms for CSC enhancement: (1) sustained executive leadership engagement; (2) context-sensitive communication strategies; and (3) policy-practice alignment. This work constitutes the first empirical validation of CSC enablers and barriers within authentic industrial settings, thereby bridging a critical research void in safety-critical domains. It delivers a theoretically grounded, operationally actionable framework to foster resilient and sustainable cybersecurity culture.

0 citationsRead paper

Gaussian Process Surrogate Models for Efficient Estimation of Structural Response Distributions and Order Statistics

Mar 03, 2025

In structural serviceability limit state (SLS) assessment under vast meteorological scenarios, efficient estimation of response distributions and order statistics—e.g., the 100th-highest response (Y_{100})—remains challenging. This paper proposes a novel Gaussian process (GP)-based surrogate modeling approach that directly treats structural response as a stochastic process. By embedding the response’s probabilistic structure into the GP, the method enables analytical generation of the full response distribution and closed-form estimation of arbitrary-order order statistics—bypassing conventional Monte Carlo resampling and costly high-fidelity simulations. Integrating uncertainty quantification with finite-sample surrogate learning, the framework achieves comparable (Y_{100}) estimation accuracy to full-physics simulation using less than 1% of its computational cost, as validated on a 25-year historical meteorological dataset. The approach significantly enhances both efficiency and scalability of structural reliability assessment under extreme weather conditions.

0 citationsRead paper