Institution profile

North Carolina State University

Academic institutionnorthamerica · us
Official website
Research library705linked papers
Opportunities0open roles
Selected work

Representative Papers

Attacking the Spike: On the Transferability and Security of Spiking Neural Networks to Adversarial Examples

Sep 07, 2022

This work systematically investigates the underexplored problem of adversarial robustness in Spiking Neural Networks (SNNs). We find that white-box attacks against SNNs heavily rely on surrogate gradient techniques and exhibit severely limited transferability of adversarial examples across architectures (e.g., between SNNs and ViTs/CNNs). To address this, we first uncover a strong coupling mechanism between SNNs’ adversarial vulnerability and surrogate gradient estimation. Building on this insight, we propose Auto-SAGA—a cross-architecture universal white-box attack method that jointly optimizes adaptive self-attention gradient estimation and surrogate gradient approximation. Evaluated on CIFAR-10, CIFAR-100, and ImageNet, Auto-SAGA achieves a 91.1% improvement in attack success rate on SNN-ViT ensembles and attains three times the effectiveness of Auto-PGD on adversarially trained SNN ensembles, significantly outperforming existing baselines.

13 citationsRead paper

No Vulnerability Data, No Problem: Towards Predicting Mean Time To Remediate In Open Source Software Dependencies

Mar 26, 2024

Assessing timely remediation risk for open-source software dependencies is challenging when vulnerability data are unavailable. Method: This paper introduces MTTR<sub>dep</sub>—Mean Time to Remediate for dependencies—and proposes MTTU<sub>dep</sub> (Mean Time to Update) as a highly correlated proxy metric, defined as the average interval between successive version releases. MTTU<sub>dep</sub> addresses key limitations of existing metrics: neglect of dependency hierarchy, insufficient historical aggregation, and insensitivity to floating-version semantics. Contribution/Results: Through large-scale empirical analysis of 163,000 packages across npm, PyPI, and Cargo ecosystems—combined with version evolution graph modeling and statistical validation—we demonstrate a strong correlation (Spearman ρ > 0.9) between MTTU<sub>dep</sub> and MTTR<sub>dep</sub>. This enables scalable, low-cost, data-light risk assessment in absence of vulnerability records, advancing quantifiable decision support for software supply chain security.

2 citationsRead paper

An efficient Monte Carlo method for valid prior-free possibilistic statistical inference

Jan 17, 2025

Traditional Monte Carlo methods struggle to efficiently approximate non-probabilistic credibility measures in possibility-based inference. Method: This paper proposes a novel inferential model (IM) framework that requires no prior specification and unifies frequentist reliability with Bayesian-like belief representation. It establishes, for the first time, a theoretical characterization of credible sets for possibility-based IMs and derives their optimal probabilistic approximation—a class of mixture distributions amenable to efficient sampling. A dedicated Monte Carlo algorithm is then designed to rapidly approximate possibility outputs with guaranteed calibration consistency and controllable error. Results: Numerical experiments demonstrate substantial improvements in both computational efficiency and accuracy: the method achieves small approximation error, fast convergence rates, and excellent frequentist calibration. It constitutes the first solution for possibility-based statistical inference that simultaneously ensures theoretical rigor and computational feasibility.

1 citations1 influentialRead paper

A Generalized Apprenticeship Learning Framework for Capturing Evolving Student Pedagogical Strategies

Feb 24, 2026International Conference on Artificial Intelligence in Education

This work addresses the challenges of low sample efficiency and reward function design in deep reinforcement learning for intelligent tutoring systems, which hinder effective modeling of students’ dynamically evolving learning strategies. To overcome these limitations, the authors propose THEMES, a framework based on generalized apprenticeship learning that introduces, for the first time, a time-varying multidimensional reward function. By leveraging only 18 expert demonstration trajectories from historical semester data, THEMES jointly integrates inverse reinforcement learning and policy generalization to accurately capture the temporal complexity and non-stationarity of teaching strategies. Evaluated on the task of predicting subsequent-semester instructional decisions, the method achieves an AUC of 0.899 and a Jaccard index of 0.653, significantly outperforming six state-of-the-art baselines and demonstrating strong efficacy and generalization capability.

1 citationsRead paper

Towards Verifiably Safe Tool Use for LLM Agents

Jan 12, 2026

This work addresses the security risks posed by large language model (LLM) agents during tool invocation, such as inadvertent leakage of sensitive data or overwriting of critical records—hazards for which existing approaches lack verifiable guarantees. To bridge this gap, the paper introduces a novel integration of System-Theoretic Process Analysis (STPA) with formal specifications to systematically identify hazards in agent workflows and derive enforceable safety requirements. These requirements are then translated into executable constraints on data flows and tool invocation sequences. Building upon an enhanced Model Context Protocol (MCP) framework, the approach incorporates structured capability control and trust-labeling mechanisms to enable proactive, verifiable protection of tool interactions. By significantly reducing reliance on manual verification, this method advances LLM agent design from empirical reliability toward a paradigm grounded in formal security assurances.

1 citationsRead paper
Recent publications

Latest Papers