2931 papers
This paper proposes SPAT-UAD, a robust fake information detection framework that integrates semantics-preserving adversarial training with uncertainty-aware adaptive decision-making to significantly outperform existing baselines in handling surface-level variations and improving model reliability across multiple datasets.
This paper introduces Symbolic Log Shield, a neurosymbolic framework that significantly mitigates the vulnerability of Large Language Models to adversarial jailbreak attacks in network log classification, effectively restoring and even enhancing their anomaly detection performance from severely degraded levels.
This paper presents a data-driven framework for LAPAN-IPB LEO satellites that combines a Hybrid SARIMAX-LSTM voltage predictor and a Random Forest classifier to generate auditable, evidence-based mission recommendations, thereby replacing reliance on undocumented operator experience with a reproducible, objective scheduling process.
The paper proposes MRT, a scene text recognition framework that combines Neighborhood Attention-enhanced Mamba for efficient local spatial aggregation with a weight-shared Reusable Transformer for bidirectional global context, achieving superior accuracy-efficiency trade-offs across multiple benchmarks.
This systematic review analyzes federated learning architectures, applications, and challenges for distributed cyber threat detection, highlighting the shift toward decentralized models to address privacy and bandwidth limitations while identifying non-IID data and adversarial attacks as critical barriers to effective implementation.
This paper proposes a fast one-step multi-view clustering method that unifies spectral clustering and nonnegative matrix factorization with tensor log-determinant regularization to effectively capture high-order cross-view correlations and achieve superior performance and scalability compared to state-of-the-art methods.
This paper proposes an REAE-LTransformer model optimized by an Improved Four-Vector Intelligent Metaheuristic (IFVIM) algorithm to accurately predict lithium-ion battery remaining useful life by effectively extracting degradation features, capturing long-term dependencies, and minimizing prediction errors, achieving superior performance with an MAE under 0.024 and R² exceeding 0.99 across multiple datasets.
This paper proposes a lightweight, XGBoost-based appearance-guided association module that integrates color, gradient, and morphological descriptors into the SORT framework to significantly reduce identity switches and improve tracking accuracy for densely stocked fish in real-time underwater monitoring.
This paper proposes a scalable hybrid encryption framework that combines a password-derived Logistic Map for chaotic key generation, a dynamic Unimodular Hill Cipher for constructing large invertible matrices, and AES-CBC to provide robust, format-independent security for universal binary files with near-ideal entropy and strong resistance to cryptanalytic attacks.
Pivot-WFSM introduces a memory-scalable approach to weighted frequent subgraph mining that replaces traditional embedding storage with on-demand re-matching, drastically reducing peak memory usage and enabling the analysis of large multigraph databases that previously caused out-of-memory failures.