Test out our new Bento Search

Small‐scale fisheries often lack historical shark and ray catch information, hampering their management. We reconstructed historical catch trends and current fishing pressure by combining local ecological knowledge, satellite‐based vessel counts, and a short‐term landing‐site survey. To test the effectiveness of this method, we focused on the Bijagós Archipelago (Guinea‐Bissau, West Africa), where historical fisheries data are lacking. Benthic rays (stingrays [Dasyatidae] and butterfly rays [Gymnura spp.]), benthopelagic rays (duckbill eagle rays [Aetomylaeus bovinus] and cownose rays [Rhinoptera marginata]), guitarfish (Glaucostegus and Rhinobatos spp.), requiem sharks (Carcharhinidae), and hammerhead sharks (Sphyrna spp.) declined in abundance by 81.5–96.7% (species dependent) from 1960 to 2020. Fishing effort increased annually: fishing trip duration by 42.0% (SE 3.4), numbers of fishing vessels at sea as perceived by fishers by 36.3% (1.0) (1960–2020), and number of vessels by 12.0% (1.1) (2007–2022). We estimated that in 2020, fishing vessels collectively captured 61–264 sharks and 522–2194 rays per day in the archipelago, depending on the proportion of the fishing fleet that was active (i.e., low fleet activity of 18% and high fleet activity of 80%). We advocate for reducing shark and ray catches by regulating fleet size, reinforcing boundaries of protected areas, and collecting fisher‐dependent information on shark and ray landings to safeguard these vulnerable species and coastal livelihoods. We demonstrated the effectiveness of using this 3‐pronged approach to provide baseline data on shark fisheries, a common challenge in areas with small‐scale fisheries and limited research capacity. [ABSTRACT FROM AUTHOR]

The U.S. imposed sanctions on China focusing on cutting-edge technology. This study examines whether the US sanctions policy against China is effective. If it is effective, we would like to analyze why it is effective through the technical diffusion index and catch-up index. As the U.S. imposed sanctions on Chinese semiconductors and other cutting-edge technologies, the speed of catch-up and diffusion of China's cutting-edge technologies slowed down. From 1980 to 2022, macro data was collected from the IMF (2023) and ILO (2023). Approximately 300,000 patents related to semiconductors, AI, and Big data were used. The impact of advanced technical diffusion on GDP and per capita income was analyzed statistically rigorously using the general sequential logit model (VGAM, VGLM), difference-in-differences (DID) model, hierarchical mixed model, and Long-Short. -Term Memory (LSTM) model. We used LSTM to assess whether China could overtake the United States by 2030. In conclusion, U.S.'s sanction on China were partially effective. It appears that China is rapidly approaching the U.S. through the development of AI and Big data technologies. However, it is doubtful whether China will be able to catch up with the U.S. by 2030. If the U.S. cannot effectively control AI, the possibility of being overtaken by China may increase.

Long-range (LoRa) wireless networks have been widely proposed as efficient wireless access networks for battery-constrained Internet of Things (IoT) devices. However, applying the LoRa-based IoT network in search-and-rescue (SAR) operations will have limited coverage caused by high signal attenuation due to terrestrial blockages, especially in highly remote areas. To overcome this challenge, using unmanned aerial vehicles (UAVs) as a flying LoRa gateway to transfer messages from ground LoRa nodes to the ground rescue station can be a promising solution. In this paper, an artificial intelligence-empowered SAR operation framework using a UAV-assisted LoRa network in different unknown search environments is designed and implemented. The problem of the flying LoRa (FL) gateway control policy is modeled as a partially observable Markov decision process to move the UAV towards the LoRa transmitter carried by a lost person in the known remote search area. A deep reinforcement learning (RL)-based policy is designed to determine the adaptive FL gateway trajectory in a given search environment. Then, as a general solution, a deep meta-RL framework is used for SAR in any new and unknown environments. The proposed deep meta-RL framework integrates the information of the prior FL gateway experience in the previous SAR environments to the new environment and then rapidly adapts the UAV control policy model for SAR operation in a new and unknown environment. To analyze the performance of the proposed framework in real-world scenarios, the proposed SAR system is experimentally tested in three environments: a university campus, a wide plain, and a slotted canyon at Mongasht mountain ranges, Iran. Experimental results show that if the deep meta-RL-based control policy is applied instead of the deep RL-based one, the number of SAR time slots decreases from 141 to 50. Moreover, in the slotted canyon environment, the UAV energy consumption under the deep meta-RL policy is respectively 57% and 23% less than the deep RL and Actor-Critic RL policies.

Phishing attacks are a major cybersecurity threat that exploit human weaknesses to steal sensitive information. Although detection systems have improved, phishing attacks remain widespread which highlights the need for human-centered defenses. In this study, we investigate the effectiveness of visual cues and explanatory feedback in improving users’ ability to detect phishing emails. We conduct a user study with 55 participants and first evaluate detection accuracy across two tasks: one without visual cues (task-1) and one with visual cues (task-2), to assess their impact. Upon completion of task-2, the participants receive explanatory feedback. Then they participate in a third task (task-3) to evaluate new emails (without visual cues). Through measuring their accuracy, we examine how email categories influence detection performance and how this changes with visual cues and feedback. Results show that visual cues offer only modest improvements in phishing detection with no statistically significant accuracy gains (task-1: 58.18% vs. task-2: 68.73%), and lead to increased misclassification of legitimate emails. In contrast, explanatory feedback significantly enhances both phishing and legitimate email detection, with overall accuracy improving to 70% in task-3, a gain of 13.82% compared to task-1 and 12.91% compared to task-2, with validated gains confirmed through a follow-up user study. Additionally, detection performance varies across email categories, with emotionally manipulative and deceptive tactics proving more difficult to detect but showing improvement after receiving explanatory feedback. These findings offer actionable insights for designing effective, feedback-based user training programs to complement automated phishing detection systems.

The wide deployment of Large Language Models (LLMs) has given rise to strong demands for optimizing their inference performance. Today’s techniques serving this purpose primarily focus on reducing latency and improving throughput through algorithmic and hardware enhancements, while largely overlooking their privacy side effects, particularly in a multi-user environment. In our research, for the first time, we discovered a set of new timing side channels in LLM systems, arising from shared caches and GPU memory allocations, which can be exploited to infer both confidential system prompts and those issued by other users. These vulnerabilities echo security challenges observed in traditional computing systems, highlighting an urgent need to address potential information leakage in LLM serving infrastructures. In this paper, we report novel attack strategies designed to exploit such timing side channels inherent in LLM deployments, specifically targeting the Key-Value (KV) cache and semantic cache widely used to enhance LLM inference performance. Our approach leverages timing measurements and classification models to detect cache hits, allowing an adversary to infer private prompts with high accuracy. We also propose a token-by-token search algorithm to efficiently recover shared prompt prefixes in the caches, showing the feasibility of stealing system prompts and those produced by peer users. Our experimental studies on black-box testing of popular online LLM services demonstrate that such privacy risks are completely realistic, with significant consequences. Our findings underscore the need for robust mitigation to protect LLM systems against such emerging threats.

Objectives: This study aimed to examine the association between weekend catch-up sleep (CUS) and obesity among Korean workers. Methods: Data were derived from the 2016–2023 Korean National Health and Nutrition Examination Survey (KNHANES), a nationally representative dataset. The final analytic sample comprised 17,208 Korean workers aged 26 to 64 years. General and abdominal obesity were defined as body mass index (BMI) ≥ 25 kg/m2 and waist circumference ≥ 90 cm for men and ≥85 cm for women, respectively. Sleep patterns were categorized into sufficient sleep, weekend CUS, and insufficient sleep. Multivariable logistic regression analyses were performed to evaluate associations between sleep patterns and obesity, adjusting for demographic, socioeconomic, and health-related variables. Results: Compared to individuals with sufficient sleep, those with weekend CUS showed increased odds of general obesity (adjusted odds ratio [AOR] = 1.21) and abdominal obesity (AOR = 1.18). The insufficient sleep group had even higher odds for both general obesity (AOR = 1.23) and abdominal obesity (AOR = 1.33). Conclusions: Insufficient sleep is significantly associated with increased risks of both general and abdominal obesity among Korean workers. While weekend CUS may offer partial mitigation of obesity risk, it should not be considered a substitute for regular, adequate sleep. Longitudinal studies are warranted to further explore causal relationships between sleep patterns and obesity in working populations. [ABSTRACT FROM AUTHOR]