Selected work

Representative Papers

Orchestrated Vulnerability Management for Heterogeneous Networks: Adaptive Two-Stage Vulnerability Assessment, Context-Aware Risk Prioritization, and Automated Mitigation

Aug 08, 2026

This study addresses the inefficiencies and delayed response in vulnerability management within heterogeneous networks, stemming from device diversity, environmental fragility, and configuration disparities. To tackle these challenges, the authors propose an automated vulnerability management framework orchestrated via SOAR (Security Orchestration, Automation, and Response). The framework integrates passive asset discovery, an adaptive two-stage vulnerability assessment, context-aware risk prioritization—combining CVSS, EPSS, and asset-specific context—and an SDN-driven mitigation mechanism capable of millisecond-level automated response. Its key innovation lies in significantly reducing scanning-induced disruption to resource-constrained devices while enabling precise risk-based prioritization. Experimental results demonstrate that the system identifies 71% of baseline vulnerabilities, reduces total scanning time by up to 91%, decreases the number of vulnerabilities requiring urgent remediation by approximately 75%, shortens assessment time for 32 hosts by up to 45%, and executes mitigation policies automatically within milliseconds.

0 citationsRead paper

Capacity of Additive-Noise Sticky Channels

Aug 02, 2026

This study addresses the long-standing open problem of determining the capacity of the sticky channel with additive noise, which preserves run-length structure under synchronization errors yet has lacked a closed-form capacity expression. By reframing the problem as a memoryless channel capacity per unit cost over the non-negative integers, and leveraging information-theoretic analysis, the capacity-per-unit-cost framework, and run-length-constrained coding, the authors derive the exact capacity for Bernoulli noise parameter \( p \leq 1/2 \). They establish that the capacity remains constant over \( p \in [1/\varphi^2, 1/2] \), fully characterizing the capacity function in this interval, and provide tight analytical bounds for \( 1/2 < p < 1 \). The approach is further extended to general additive noise distributions, precisely identifying parameter regimes where zero-error run-length-constrained codes are suboptimal.

0 citationsRead paper

Diffusion Models in Medical Image Inpainting: Challenges, Solution Taxonomy, and Future Directions

Jul 23, 2026

Medical images are frequently compromised by artifacts, missing regions, or pathological alterations, which can undermine diagnostic reliability. This work presents a systematic review of diffusion model–based approaches for medical image inpainting and introduces the first taxonomy specifically tailored to this domain. The proposed framework encompasses prevailing architectures—such as Denoising Diffusion Probabilistic Models (DDPM) and Latent Diffusion Models (LDM)—alongside key clinical applications (e.g., MRI and CT), benchmark datasets, and evaluation protocols. Empirical analysis demonstrates that diffusion models excel at generating anatomically plausible reconstructions, yet critical challenges persist, notably the absence of standardized benchmarks and limited data diversity. By synthesizing current advances and identifying open problems, this study offers a structured foundation to guide future research in medical image restoration.

0 citationsRead paper

The Insertion List-Decoding Capacity and an Improved Bound on the Deletion List-Decoding Capacity

Jul 04, 2026

This work investigates the list-decoding capacity of binary codes under synchronization errors caused by insertions and deletions, with a focus on clarifying capacity bounds in high-error regimes where they remain poorly understood. By constructing random codes via a symmetric two-state Markov chain and combining information-theoretic and combinatorial coding arguments, the authors precisely characterize the list-decoding capacity for any insertion fraction δ ∈ [0,1] as (1+δ)(1−h(δ/(1+δ))), where h(·) denotes the binary entropy function. They further demonstrate that such codes outperform uniformly random codes over insertion channels. Additionally, the study derives a tighter upper bound on the deletion channel capacity, which asymptotically matches the known binary deletion channel capacity 1−h(δ) as δ→0.

0 citationsRead paper

MultiSynt/MT: Trillion-Token Multi-Parallel Pre-Training Data Translated Across 36 Languages

Jul 01, 2026

This work addresses the scarcity of non-English open-source pretraining corpora, which hinders the development of multilingual large language models. To overcome this limitation, the authors propose a machine translation–based synthesis approach that leverages the high-quality Nemotron-CC corpus and employs both Tower+ and OPUS-MT/HPLT-MT systems to generate a sentence-aligned parallel corpus spanning 36 European languages and approximately 4.8 trillion tokens—the first large-scale, open, multi-system-fused multilingual pretraining dataset of its kind. Experimental results demonstrate that, under a fixed budget of 100 billion tokens, models trained on this synthetic data achieve a ~15% performance gain over the native-data baseline HPLT 2.0; moreover, they reach equivalent final performance using only 72% of the token budget, substantially reducing reliance on scarce native multilingual data and exposing limitations in current evaluation benchmarks.

0 citationsRead paper

Recent publications

Latest Papers

Orchestrated Vulnerability Management for Heterogeneous Networks: Adaptive Two-Stage Vulnerability Assessment, Context-Aware Risk Prioritization, and Automated Mitigation

Aug 08, 2026

This study addresses the inefficiencies and delayed response in vulnerability management within heterogeneous networks, stemming from device diversity, environmental fragility, and configuration disparities. To tackle these challenges, the authors propose an automated vulnerability management framework orchestrated via SOAR (Security Orchestration, Automation, and Response). The framework integrates passive asset discovery, an adaptive two-stage vulnerability assessment, context-aware risk prioritization—combining CVSS, EPSS, and asset-specific context—and an SDN-driven mitigation mechanism capable of millisecond-level automated response. Its key innovation lies in significantly reducing scanning-induced disruption to resource-constrained devices while enabling precise risk-based prioritization. Experimental results demonstrate that the system identifies 71% of baseline vulnerabilities, reduces total scanning time by up to 91%, decreases the number of vulnerabilities requiring urgent remediation by approximately 75%, shortens assessment time for 32 hosts by up to 45%, and executes mitigation policies automatically within milliseconds.

0 citationsRead paper

Capacity of Additive-Noise Sticky Channels

Aug 02, 2026

This study addresses the long-standing open problem of determining the capacity of the sticky channel with additive noise, which preserves run-length structure under synchronization errors yet has lacked a closed-form capacity expression. By reframing the problem as a memoryless channel capacity per unit cost over the non-negative integers, and leveraging information-theoretic analysis, the capacity-per-unit-cost framework, and run-length-constrained coding, the authors derive the exact capacity for Bernoulli noise parameter \( p \leq 1/2 \). They establish that the capacity remains constant over \( p \in [1/\varphi^2, 1/2] \), fully characterizing the capacity function in this interval, and provide tight analytical bounds for \( 1/2 < p < 1 \). The approach is further extended to general additive noise distributions, precisely identifying parameter regimes where zero-error run-length-constrained codes are suboptimal.

0 citationsRead paper

Diffusion Models in Medical Image Inpainting: Challenges, Solution Taxonomy, and Future Directions

Jul 23, 2026

Medical images are frequently compromised by artifacts, missing regions, or pathological alterations, which can undermine diagnostic reliability. This work presents a systematic review of diffusion model–based approaches for medical image inpainting and introduces the first taxonomy specifically tailored to this domain. The proposed framework encompasses prevailing architectures—such as Denoising Diffusion Probabilistic Models (DDPM) and Latent Diffusion Models (LDM)—alongside key clinical applications (e.g., MRI and CT), benchmark datasets, and evaluation protocols. Empirical analysis demonstrates that diffusion models excel at generating anatomically plausible reconstructions, yet critical challenges persist, notably the absence of standardized benchmarks and limited data diversity. By synthesizing current advances and identifying open problems, this study offers a structured foundation to guide future research in medical image restoration.

0 citationsRead paper

The Insertion List-Decoding Capacity and an Improved Bound on the Deletion List-Decoding Capacity

Jul 04, 2026

This work investigates the list-decoding capacity of binary codes under synchronization errors caused by insertions and deletions, with a focus on clarifying capacity bounds in high-error regimes where they remain poorly understood. By constructing random codes via a symmetric two-state Markov chain and combining information-theoretic and combinatorial coding arguments, the authors precisely characterize the list-decoding capacity for any insertion fraction δ ∈ [0,1] as (1+δ)(1−h(δ/(1+δ))), where h(·) denotes the binary entropy function. They further demonstrate that such codes outperform uniformly random codes over insertion channels. Additionally, the study derives a tighter upper bound on the deletion channel capacity, which asymptotically matches the known binary deletion channel capacity 1−h(δ) as δ→0.

0 citationsRead paper

MultiSynt/MT: Trillion-Token Multi-Parallel Pre-Training Data Translated Across 36 Languages

Jul 01, 2026

This work addresses the scarcity of non-English open-source pretraining corpora, which hinders the development of multilingual large language models. To overcome this limitation, the authors propose a machine translation–based synthesis approach that leverages the high-quality Nemotron-CC corpus and employs both Tower+ and OPUS-MT/HPLT-MT systems to generate a sentence-aligned parallel corpus spanning 36 European languages and approximately 4.8 trillion tokens—the first large-scale, open, multi-system-fused multilingual pretraining dataset of its kind. Experimental results demonstrate that, under a fixed budget of 100 billion tokens, models trained on this synthetic data achieve a ~15% performance gain over the native-data baseline HPLT 2.0; moreover, they reach equivalent final performance using only 72% of the token budget, substantially reducing reliance on scarce native multilingual data and exposing limitations in current evaluation benchmarks.

0 citationsRead paper