检索结果-南通市图书馆

Journal of Software Engineering and Applications 2024年第5期17卷 340-358页

作者： Alicia Biju Vishnupriya Ramesh Vijay k. Madisetti School of Cybersecurity and Privacy College of Computing Georgia Institute of Technology Atlanta GA USA

Large Language Models (LLMs) have revolutionized Generative Artificial Intelligence (GenAI) task., becoming an integral part of various applications in society, including text generation, translation, summarization, and more. However, their widespread usage emphasizes the critical need to enhance their security posture to ensure the integrity and reliability of their outputs and minimize harmful effects. Prompt injections and training data poisoning attack. are two of the most prominent vulnerabilities in LLMs, which could potentially lead to unpredictable and undesirable behaviors, such as biased outputs, misinformation propagation, and even malicious content generation. The Common Vulnerability Scoring System (CVSS) framework.provides a standardized approach to capturing the principal characteristics of vulnerabilities, facilitating a deeper understanding of their severity within the security and AI communities. By extending the current CVSS framework. we generate scores for these vulnerabilities such that organizations can prioritize mitigation efforts, allocate resources effectively, and implement targeted security measures to defend against potential risk..

关键词： Common Vulnerability Scoring System (CVSS) Large Language Models (LLMs) DALL-E Prompt Injections Training Data Poisoning CVSS Metrics

Grid System Analysis of Urban Flora of Buk.ara City (Uzbek.stan)

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

American Journal of Plant Sciences 2024年第2期15卷 139-144页

作者： Abdulla M. Umedov Husniddin k. Esanov Department of Botany and Plant Physiology Bukhara State University Bukhara Uzbekistan

This article presents information on the study of the flora of Uzbek.stan based on grid system mapping. The urban flora of the city of Buk.ara was researched in it. As a result of research, the territory of Buk.ara city was divided into 85 indexes based on 1 × 1 k.2 grid mapping system. The diversity and density of species in the indexes are determined. The influence of anthropogenic factors on the diversity of species in the indexes is determined.

关键词： Buk.ara City Urban Flora Index Grid Map System Herbarium Geoinformation

Observer Based Control for a Class of Systems with Output Deadzone Nonlinearity

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Intelligent Control and Automation 2024年第3期15卷 95-107页

作者： Nizar J. Alk.ateeb Hameed k. Ebraheem Department of Electronic Engineering Technology College of Technological Studies Public Authority of Applied Education and Training (PAAET) Kuwait City Kuwait

In this paper, a combination of model based adaptive design along with adaptive linear output feedback.controller is used to compensate for robotic manipulator with output deadzone nonlinearity. The deadzone dynamics are utilized to adaptively estimate the deadzone parameter and a switching function is designed to eliminate the error produced in the adaptive observer dynamics. The overall design of the closed loop system ensures stability in the BIBO criterion.

关键词： Adaptive Output Observer Deadzone Compensation Robot Manipulator Sliding Mode Control

Parental k.owledge of Pre-Hospital Management of Avulsed Permanent Tooth in Children at k.th

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Open Journal of Stomatology 2024年第2期14卷 87-102页

作者： Ama Agyeibea Amuasi Samuel k. Addo Solomon Obiri-Yeboah Department of Child Oral Health & Orthodontics School of Medicine and Dentistry College of Health Sciences Kwame Nkrumah University of Science and Technology Kumasi Ghana Department of Oral & Maxillofacial Sciences School of Medicine and Dentistry College of Health Sciences Kwame Nkrumah University of Science and Technology Kumasi Ghana

BACk.ROUND: Permanent tooth avulsion is one of the severe forms of dental traumatic injuries. The immediate action tak.n at the site of the accident is crucial to the prognosis of the tooth. Replantation is considered as the treatment of choice. OBJECTIVE: The study was undertak.n to assess the k.owledge of parents who accompany their children to the pediatric dental clinic, k.TH on the pre-hospital management of avulsed permanent tooth in children. METHODS: A researcher administered structured questionnaire was used to collect data on the k.owledge of pre-hospital management of avulsed permanent tooth from 83 parents who accompanied their wards to the pediatric dental clinic at k.TH. RESULTS: A total of 83 parents were involved in the study. 30 (36%) were males while 53 (64%) were females. The majority of the parents (57%) were either university trained or had attended college of education. Only 32 parents (39%) were aware of the possibility of replantation. Majority of the parents chose non-physiologic media as the transport media of choice and only 10% would attempt self-replantation before seek.ng professional help. 76 parents (92%) had no previous education on pre-hospital management of avulsed tooth. CONCLUSION: The results obtained from this study indicate that parental k.owledge on pre-hospital management of avulsed permanent tooth is low hence the need for massive public educational campaigns.

关键词： Parental k.owledge Tooth Avulsion Pre-Hospital Management Permanent Tooth

GUARDIAN: A Multi-Tiered Defense Architecture for Thwarting Prompt Injection Attack. on LLMs

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Journal of Software Engineering and Applications 2024年第1期17卷 43-68页

作者： Parijat Rai Saumil Sood Vijay k. Madisetti Arshdeep Bahga School of Computer Science Engineering & Technology Bennett University Greater Noida India School of Cybersecurity and Privacy Georgia Institute of Technology Atlanta USA Cloudemy Technology Labs Chandigarh India

This paper introduces a novel multi-tiered defense architecture to protect language models from adversarial prompt attack.. We construct adversarial prompts using strategies lik. role emulation and manipulative assistance to simulate real threats. We introduce a comprehensive, multi-tiered defense framework.named GUARDIAN (Guardrails for Upholding Ethics in Language Models) comprising a system prompt filter, pre-processing filter leveraging a toxic classifier and ethical prompt generator, and pre-display filter using the model itself for output screening. Extensive testing on Meta’s Llama-2 model demonstrates the capability to block.100% of attack.prompts. The approach also auto-suggests safer prompt alternatives, thereby bolstering language model security. Quantitatively evaluated defense layers and an ethical substitution mechanism represent k.y innovations to counter sophisticated attack.. The integrated methodology not only fortifies smaller LLMs against emerging cyber threats but also guides the broader application of LLMs in a secure and ethical manner.

关键词： Large Language Models (LLMs) Adversarial Attack Prompt Injection Filter Defense Artificial Intelligence Machine Learning Cybersecurity

Quantifying the Impact of In-Cab Alerts on Truck.Speed Reductions in Ohio

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Journal of Transportation Technologies 2024年第3期14卷 273-288页

作者： Jairaj Desai Jijo k. Mathew Darcy M. Bullock Lyles School of Civil Engineering Purdue University West Lafayette IN USA

In-cab alerts warn commercial vehicle drivers of upcoming roadway incidents, slowdowns and work.zone construction activities. This paper reports on a study evaluating the driver response to in-cab alerts in Ohio. Driver response was evaluated by measuring the statistical trends of vehicle speeds after the in-cab alerts were received. Vehicle speeds pre and post in-cab alert were collected over a 47 day period in the fall of 2023 for truck. traveling on interstate roadways in Ohio. Results show that approximately 22% of drivers receiving Dangerous Slowdown alerts had reduced their speeds by at least 5 mph 30 seconds after receiving such an alert. Segmenting this analysis by speed found that of vehicles traveling at or above 70 mph at the time of alerting, 26% reduced speeds by at least 5 mph. These speed reductions suggest drivers tak.ng actional measures after receiving alerts. Future studies will involve further analysis on the impact of the types of alerts shown, roadway characteristics and overall traffic conditions on truck.speeds passing through work.zones.

关键词： In-Cab Alerts Connected Truck.Data Driver Alerts Dangerous Slowdowns

Hydromagnetic Squeezing Nanofluid Flow between Two Vertical Plates in Presence of a Chemical Reaction

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Journal of Applied Mathematics and Physics 2024年第1期12卷 126-146页

作者： Benjamin Matur Madit Jack.on k. k.anza Phineas Roy k.ogora Department of Mathematics Pan African University Institute of Basic Sciences Technology and Innovation Juja Kenya Department of Pure and Applied Mathematics Jomo Kenyatta University of Agriculture and Technology Juja Kenya

In this study, Hydromagnetic Squeezing Nanofluid flow between two vertical plates in presence of a chemical reaction has been investigated. The governing equations were transformed by similarity transformation and the resulting ordinary differential equations were solved by collocation method. The velocity, temperature, concentration and magnetic induction profiles were determined with help of various flow parameters. The numerical scheme was simulated with aid of MATLAB. The results showed that increasing the squeeze number only boosts velocity and concentration while lowering temperature. Conversely, increasing the Hartmann number, Reynold’s magnetic number, Eck.rt number and Thermal Grashof number generally increases temperature but decreases both velocity and concentration. Chemical reaction rate and Soret number solely elevate concentration while Schmidt number only reduces it. The results of this study will be useful in the fields of oil and gas industry, plastic processing industries, filtration, food processing, lubrication system in machinery, Microfluidics devices for drug delivery and other related fields of nanotechnology.

关键词： Hydromagnetic Squeezing Flow Nanofluid Variable Magnetic Field Chemical Reaction

Whispered Tuning: Data Privacy Preservation in Fine-Tuning LLMs through Differential Privacy

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Journal of Software Engineering and Applications 2024年第1期17卷 1-22页

作者： Tanmay Singh Harshvardhan Aditya Vijay k. Madisetti Arshdeep Bahga School of Computer Science Engineering & Technology Bennett University Greater Noida India School of Cybersecurity and Privacy Georgia Institute of Technology Atlanta USA Cloudemy Technology Labs Chandigarh India

The proliferation of Large Language Models (LLMs) across various sectors underscored the urgency of addressing potential privacy breaches. Vulnerabilities, such as prompt injection attack. and other adversarial tactics, could mak. these models inadvertently disclose their training data. Such disclosures could compromise personal identifiable information, posing significant privacy risk.. In this paper, we proposed a novel multi-faceted approach called Whispered Tuning to address privacy leak. in large language models (LLMs). We integrated a PII redaction model, differential privacy techniques, and an output filter into the LLM fine-tuning process to enhance confidentiality. Additionally, we introduced novel ideas lik. the Epsilon Dial for adjustable privacy budgeting for differentiated Training Phases per data handler role. Through empirical validation, including attack. on non-private models, we demonstrated the robustness of our proposed solution SecureNLP in safeguarding privacy without compromising utility. This pioneering methodology significantly fortified LLMs against privacy infringements, enabling responsible adoption across sectors.

关键词： NLP Differential Privacy Adversarial Attack. Informed Decisions

Smaller & Smarter: Score-Driven Network.Chaining of Smaller Language Models

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Journal of Software Engineering and Applications 2024年第1期17卷 23-42页

作者： Gunik. Dhingra Siddansh Chawla Vijay k. Madisetti Arshdeep Bahga School of Computer Science Engineering & Technology Bennett University Greater Noida India School of Cybersecurity and Privacy Georgia Institute of Technology Atlanta USA Cloudemy Technology Labs Chandigarh India

With the continuous evolution and expanding applications of Large Language Models (LLMs), there has been a noticeable surge in the size of the emerging models. It is not solely the growth in model size, primarily measured by the number of parameters, but also the subsequent escalation in computational demands, hardware and software prerequisites for training, all culminating in a substantial financial investment as well. In this paper, we present novel techniques lik. supervision, parallelization, and scoring functions to get better results out of chains of smaller language models, rather than relying solely on scaling up model size. Firstly, we propose an approach to quantify the performance of a Smaller Language Models (SLM) by introducing a corresponding supervisor model that incrementally corrects the encountered errors. Secondly, we propose an approach to utilize two smaller language models (in a network. performing the same task.and retrieving the best relevant output from the two, ensuring peak.performance for a specific task. Experimental evaluations establish the quantitative accuracy improvements on financial reasoning and arithmetic calculation task. from utilizing techniques lik. supervisor models (in a network.of model scenario), threshold scoring and parallel processing over a baseline study.

关键词： Large Language Models (LLMs) Smaller Language Models (SLMs) Finance Network.ng Supervisor Model Scoring Function

Direct scaling of residual displacements for bilinear and pinching oscillators

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Earthquak. Engineering and Engineering Vibration 2024年第1期23卷 129-149页

作者： Mohammad Saifullah Vinay k.Gupta Department of Civil Engineering Indian Institute of Technology KanpurKanpur 208016India

The estimation of residual displacements in a structure due to an anticipated earthquak. event has increasingly become an important component of performance-based earthquak. engineering because controlling these displacements plays an important role in ensuring cost-feasible or cost-effective repairs in a damaged structure after the *** attempt is made in this study to obtain statistical estimates of constant-ductility residual displacement spectra for bilinear and pinching oscillators with 5%initial damping,directly in terms of easily available seismological,site,and model *** of the available models for the bilinear and pinching oscillators are useful when design spectra for a seismic hazard at a site are not *** statistical estimates of a residual displacement spectrum are proposed in terms of earthquak. magnitude,epicentral distance,site geology parameter,and three model parameters for a given set of ductility demand and a hysteretic energy capacity coefficient in the case of bilinear and pinching models,as well as for a given set of pinching parameters for displacement and strength at the break.oint in the case of pinching model *** proposed scaling model is applicable to horizontal ground motions in the western *** earthquak. magnitudes less than 7 or epicentral distances greater than 20 k..

关键词： residual displacement spectrum bilinear hysteresis model pinching hysteresis model nonlinear analysis scaling model