Eddy current testing (ECT) is a widely adopted electromagnetic non-destructive testing (NDT) technique for detecting defects in conductive materials. In practical deployments, however, ECT systems often suffer from low signal-to-noise ratio, strong sensitivity to lift-off and environmental variations, and complex multi-parameter coupling, which makes robust signal interpretation challenging. Meanwhile, the growing demand for portable and always-on inspection pushes data processing toward resource-constrained embedded hardware. Tiny machine learning (TinyML) provides a promising pathway for enabling on-device intelligence by deploying compact models with low latency and low power consumption. This review summarizes recent progress in integrating TinyML into ECT, covering the ECT signal characteristics and key technical bottlenecks, the TinyML workflow and optimization techniques for embedded deployment, and representative application scenarios including pipeline inspection, corrosion detection, and thickness evaluation. We further analyze the main barriers to adoption, such as limited computing power and memory, data scarcity, calibration overhead, and generalization across materials, probes, and defect types, and we outline future research directions including physics-guided learning, federated learning, and standardized benchmarks for ECT-oriented TinyML evaluation.

Large Language Models (LLMs) exhibit remarkable language understanding and problem-solving capabilities but are heavily reliant on vast amounts of high-quality data, which raises significant privacy concerns. Federated Learning (FL) presents a decentralized training paradigm that minimizes raw data sharing but is still vulnerable to privacy attacks that exploit shared model updates. Federated LLMs, especially dialogue models, are particularly prone to query-based attacks, which can lead to the unintentional disclosure of memorized private information. This study performs extensive experiments across multiple model sizes and federated architectures to assess the extent of query-based privacy leakage and evaluate existing protection mechanisms. To enable standardized and reproducible privacy assessments, we propose a multi-stage privacy leakage detection method, integrating structured privacy attribute extraction, controlled injection, realistic prompting, and quantitative evaluation under both generalized and personalized privacy scenarios. Additionally, we introduce FedShieldLLM, a lightweight, adaptable client-side anonymization framework that performs personalized data de-identification prior to federated training, effectively mitigating privacy risks while preserving task utility. Our contributions include: (1) a systematic analysis of query-based privacy leakage in federated LLMs; (2) the design of a reproducible, architecture-agnostic multi-stage detection method; and (3) the development of FedShieldLLM for secure and flexible client participation in privacy-sensitive federated learning.

Federated learning (FL) stands out as a promising solution to address the ever-increasing data scarcity problem of large language model (LLM) training through collecting data from distributed sources. The unique challenges however exist in three aspects, the extreme large communication overhead due to frequent model aggregation, the suboptimal performance due to data heterogeneity, and privacy leakage risk due to model parameter exchange. To address these issues, we propose Federated Knowledge Learning (FKL), a novel token-level knowledge sharing framework tailored for LLM to achieve effective task relevant knowledge aggregation. FKL offers the following advantages: (1) Adaptive Flexibility: Clients and servers can employ models customized for their specific tasks while ensuring seamless collaboration. Each node within the system functions independently and can be flexibly adapted as needed. (2) Communication Efficiency: Instead of huge model or parameter exchange, the unique knowledge sharing mechanism of FKL can significantly reduce the communication overhead while delivering superior performance. (3) Privacy Safeguards: The distributed token-level knowledge-sharing mechanism ensures privacy by deidentifying sensitive information and mitigating potential leakage risks. Experiments on LLMs (125M-7B parameters) validate FKL's versatility, achieving 50% higher accuracy on GLUE benchmarks and a notable 0.7x improvement in privacy task response quality compared to existing FL-based LLM approaches.

As essential components for communication, network protocol programs are highly security-critical, making it crucial to identify their vulnerabilities. Fuzzing is one of the most popular software vulnerability discovery techniques, being highly efficient and having low false-positive rates. However, current network protocol fuzzing is hindered by the randomness in programs. The current solutions primarily rely on the manual modification of programs, which is inefficient and prone to omissions. In this paper, we propose DerandomPre, a novel stability enhancement method for stateful network protocol programs, which leverages large language model’s code- and text-understanding capabilities to analyze derandomization knowledge and optimize the stability enhancing of programs for fuzzing. DerandomPre automatically eliminates randomness in programs to ensure higher stability and fuzzing effectiveness. We implement a prototype of DerandomPre. The evaluation demonstrates that DerandomPre significantly enhances fuzzing performance by eliminating program randomness. Specifically, compared to unmodified programs, DerandomPremodified versions achieved an average stability improvement of 64.88%; relative to ProFuzzBench-modified programs, DerandomPre yielded a further 13.08% stability gain. Moreover, DerandomPre demonstrates good scalability, thus is applicable to various network protocol programs.

The quantitative assessment of human error risk in computerized aircraft cockpit need a comprehensive understanding of performance shaping factor (PSF) effects on interaction tasks and the dependency between PSFs. This study employed stepwise logistic regression analysis and mediating effect analysis to explore PSF dependencies based on 872 air accident report data. Among the 14 PSFs of the established PSF taxonomy, five PSFs have significant influences on crew error. Further dependency analysis shows that a significant mediating effect exists in 16 PSF combinations, however, only 8 combinations should be considered in the actual human reliability analysis (HRA) process (indirect effect exceeds 20%). The study proposes an empirical method for identifying and assessing PSF dependencies. The results can be used to develop new or modify existing HRA methods according to the civil aviation context, as well as provide guidance to human error risk prevention and mitigation from broader perspectives in addition to interface facet.

The pupil diameter is one of the main indices in eye movement research. Quantitative analysis of pupil diameter can provide the possibility for in-depth study based on eye research. The aim of this article is description and validation of a method for conversion between the pupil diameter measured by eye tracker and the real value, black-round piece method. Firstly, the diameter value of black-round piece made by ourselves was measured by eye tracker and vernier caliper. After getting the value ratio of the eye movement measurement and caliper measurement, it can be used for calculating real pupil diameter by measurement value by eye tracker. In the article, the camera photography method was adopted to measuring real pupil diameter for validation the black - round piece method. Experimental results showed that all the error percent of pupil diameter from the black - round piece method and camera photography method are less than 5%, verified the availability of using black-round piece method on conversion between the measured value by eye tracker with optical recording and the real pupil diameter value.

The traditional measurement method of pilot's mental workload is subjective measurement, which is difficult to implement real-time measurement. Much work has done on the eye activity measurement for mental workload, but most of the work just found the relationship between eye movement indexes and mental workload, which did not show how to evaluate mental workload by eye activity measurement. The goal of the research is to find the application method of eye activity measurement on pilot's mental workload. The aim of this article is to search for the eye movement indices, which can reflect the mental workload in basic flight task. The button operation experiment is achieved by C programming software. There are nine experiment trials of different interval time. Subjects were required to press the keyboard key corresponding with the letter appearing on the screen. The eye indices were recorded by eye tracking. Subjective survey was conducted on time pressure and mental workload. Operation results were recorded in experimental procedures. In our experiment, the analysis of experimental results showed that pupil size, average fixation time, fixation frequency, saccade frequency and average saccade velocity changed with the mental workload significantly. No certain conclusion was obtained concerning the relation between blink frequency and mental workload. The research results will provide experimental basic for the workload measurement of flight task by eye activity movement.