检索结果-内蒙古大学图书馆

Effective and Imperceptible Adversarial Textual Attack Via Multi-objectivization

ACM Transactions on Evolutionary Learning and Optimization 2024年第3期4卷 1-23页

作者： Liu, Shengcai Lu, Ning Hong, Wenjing Qian, Chao Tang, Ke Department of Computer Science and Engineering Southern University of Science and Technology Shenzhen College of Engineering South Tower 1088 Xueyuan Avenue Guangdong Shenzhen518055 China Department of Computer Science and Engineering Hong Kong University of Science and Technology Hong Kong999077 Hong Kong National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Shenzhen 10F Zhizhen Building Canghai Campus 3688 Nanhai Avenue Guangdong Shenzhen China National Key Laboratory for Novel Software Technology School of Artificial Intelligence Nanjing University Nanjing International College Area Xianlin Campus 163 Xianlin Avenue Jiangsu Nanjing China

The field of adversarial textual attack has significantly grown over the past few years, where the commonly considered objective is to craft adversarial examples (AEs) that can successfully fool the target model. However, the imperceptibility of attacks, which is also essential for practical attackers, is often left out by previous studies. In consequence, the crafted AEs tend to have obvious structural and semantic differences from the original human-written text, making them easily perceptible. In this work, we advocate leveraging multi-objectivization to address such an issue. Specifically, we reformulate the problem of crafting AEs as a multi-objective optimization problem, where the attack imperceptibility is considered as an auxiliary objective. Then, we propose a simple yet effective evolutionary algorithm, dubbed HydraText, to solve this problem. HydraText can be effectively applied to both score-based and decision-based attack settings. Exhaustive experiments involving 44,237 instances demonstrate that HydraText consistently achieves competitive attack success rates and better attack imperceptibility than the recently proposed attack approaches. A human evaluation study also shows that the AEs crafted by HydraText are more indistinguishable from human-written text. Finally, these AEs exhibit good transferability and can bring notable robustness improvement to the target model by adversarial training. © 2024 Copyright held by the owner/author(s).

关键词： Multiobjective optimization

来源：评论

学校读者我要写书评

暂无评论

Visual Learning with Dynamic Recall 1

引用

5th International Conference on Soft computing and Signal Processing, ICSCSP 2022

作者： Revathy, G. Sree, Pokkuluri Kiran Devi, S. Sasikala Karunamoorthi, R. Vadivu, S. Senthil School of Computing SASTRA Deemed University Tamilnadu Thanjavur India W.G Dt AP VishnupurBhimavaram India School of Computer Science PPG College of Arts and Science Tamilnadu Coimbatore India Department of Computer Science and Engineering Kongu Engineering College Tamilnadu Perundurai India Department of IT Arulmigu Meenakshi Amman College of Engineering Tamilnadu Thiruvannamalai India

ISBN: (数字)9789811986697

ISBN: (纸本)9789811986680

The domain of Computer Vision is based on understanding the human visual system and replicating its functionalities using artificial systems. Existing CV algorithms have been able to emulate the human visual system through tasks which include but are not limited to image captioning, classifying visuals and recognizing objects. The human visual system’s astounding capability to recognize objects after encountering a few instances of them has not been replicated by an artificial system. This is crucial in deploying object recognition systems in a real-world scenario as it is not possible to predetermine the categories of objects that the systems will encounter. The CV task that pertains to it is called few-shot learning. Although Deep Neural Networks have attained impressive accuracies in recognizing objects when sufficiently large amounts of data are available, it is not feasible to collect the necessary data for every single object category. This project is an attempt to study and implement a few-shot learning algorithm that extends an object recognition system with a few-shot classification weight generator using knowledge about the base categories to generate weights for the novel categories which have limited samples. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Computer vision

来源：评论

学校读者我要写书评

暂无评论

Fall Detection using LiDAR Sensor with PointNet++

Fall Detection using LiDAR Sensor with PointNet++

引用

Sentiment Analysis and Deep Learning (ICSADL), International Conference on

作者： Anish Sai Ayyagari Syed Ashfaq Ahmed Shri Venkatakrishnan Senthil Kumar Thangavel Jeyakumar G Madhusudana Rao Nalluri Department of Computer Science and Engineering Amrita School of Computing Amrita Vishwa Vidyapeetham Coimbatore India

ISBN: (数字)9798331523923

ISBN: (纸本)9798331523930

In today's world, fall incidents pose a significant risk to the elderly, often resulting in severe consequences. Falls can cause damage to the body, especially in elders, where it can be *** is vital to detect falls in order to provide assistance. The simple solution is to install cameras in required areas, which would constantly monitor the area for any falls occurring, this poses a privacy issue in high-risk regions such as bathrooms and toilets. Keeping this in mind, this paper proposes the use of LiDAR sensors that do not cause any invasion of privacy for fall detection. The paper further uses the Pointnet++ to detect fall from the LiDAR scans. The model achieves an accuracy of 85.1 % and precision of 84.8 % on a 70–30 split of the dataset and an accuracy of 86.3 % and precision of 86.2 % on an 80–20 split of the dataset.

关键词： Deep learning Privacy Sentiment analysis Laser radar Accuracy Cameras Sensors Fall detection Older adults Monitoring

来源：评论

学校读者我要写书评

暂无评论

Real-Time Task Manager: A Python-Based Approach Using Psutil and Tkinter

Real-Time Task Manager: A Python-Based Approach Using Psutil...

引用

International Conference on Computation System and Information Technology for Sustainable Solutions (CSITSS)

作者： P. Charan Sai K. Karthik K. Bhargav Prasad C. Venkata Sai Pranav Divya K.V Computer Science and Engineering Amrita School of Computing Bengaluru Amrita Vishwa Vidyapeetham India

ISBN: (数字)9798331505462

ISBN: (纸本)9798331505479

The creation of a Real-Time Task Manager with a user-friendly interface and effective system monitoring is presented in this work utilising Python. The Psutil library was used to retrieve comprehensive system information, and Tkinter was used to develop the graphical user interface (GUI). The task manager's capacity to show parent processes at first, enabling users to expand and examine matching child processes upon request, is one of its primary features. In addition, the system permits the termination of user processes and facilitates the search for running processes. Real-time process data refreshing is another essential feature that makes sure the most recent information is shown. The Process ID, Process Name, Username, Memory Usage Percentage, and CPU Usage Percentage are among the prominent metrics displayed. These indicators are essential for locating operations that use a lot of resources and for efficiently controlling system performance. Iterative development was prioritised in the design process, including user feedback to improve stability and performance. The task manager is a potent tool for users and system administrators, providing real-time updates and responsive monitoring and management features through its modular and scalable architecture.

关键词： Measurement System performance Real-time systems User experience Libraries Iterative methods Information technology Monitoring Thermal stability Graphical user interfaces

来源：评论

学校读者我要写书评

暂无评论

Fine-Grained Multivariate Time Series Anomaly Detection in IoT

引用

Computers, Materials & Continua 2023年第6期75卷 5027-5047页

作者： Shiming He Meng Guo Bo Yang Osama Alfarraj Amr Tolba Pradip Kumar Sharma Xi’ai Yan School of Computer&Communication Engineering Changsha University of Science&TechnologyChangsha410114China Computer Science Department Community CollegeKing Saud UniversityRiyadh11437Saudi Arabia Department of Computing Science University of AberdeenAberdeenAB243FXUK Hunan Provincial Key Laboratory of Network Investigational Technology Hunan Police AcademyChangsha410138China

Sensors produce a large amount of multivariate time series data to record the states of Internet of Things(IoT)*** time series timestamp anomaly detection(TSAD)can identify timestamps of attacks and ***,it is necessary to determine which sensor or indicator is abnormal to facilitate a more detailed diagnosis,a process referred to as fine-grained anomaly detection(FGAD).Although further FGAD can be extended based on TSAD methods,existing works do not provide a quantitative evaluation,and the performance is ***,to tackle the FGAD problem,this paper first verifies that the TSAD methods achieve low performance when applied to the FGAD task directly because of the excessive fusion of features and the ignoring of the relationship’s dynamic changes between ***,this paper proposes a mul-tivariate time series fine-grained anomaly detection(MFGAD)*** avoid excessive fusion of features,MFGAD constructs two sub-models to independently identify the abnormal timestamp and abnormal indicator instead of a single model and then combines the two kinds of abnormal results to detect the fine-grained *** on this framework,an algorithm based on Graph Attention Neural Network(GAT)and Attention Convolutional Long-Short Term Memory(A-ConvLSTM)is proposed,in which GAT learns temporal features of multiple indicators to detect abnormal timestamps and A-ConvLSTM captures the dynamic relationship between indicators to identify abnormal *** simulations on a real-world dataset demonstrate that the proposed algorithm can achieve a higher F1 score and hit rate than the extension of existing TSAD methods with the benefit of two independent sub-models for timestamp and indicator detection.

关键词： Multivariate time series graph attention neural network fine-grained anomaly detection

来源：评论

学校读者我要写书评

暂无评论

Empowering Collaborative Filtering Using Probabilistic Diffusion Models

Empowering Collaborative Filtering Using Probabilistic Diffu...

引用

International Conference on Information and Communication Technology Convergence (ICTC)

作者： Yu Hou Won-Yong Shin School of Mathematics and Computing (Computational Science and Engineering) Yonsei University Seoul Republic of Korea

ISBN: (数字)9798350364637

ISBN: (纸本)9798350364644

Probabilistic diffusion models, a type of deep gen-erative models, have become one of emerging topics thanks to their state-of-the-art performance in computer vision and natural language processing (NLP) domains. Due to their denoising nature, diffusion models align well with recommender systems, where noisy user-item historical interactions are given. Diffusion models can effectively recover the original interactions from corrupted ones. Nevertheless, there are various challenges in applying diffusion models to recommender systems. To under-stand the challenges and solutions, there should be a reference for researchers and practitioners working on diffusion-based recommender systems. To this end, we first provide a comprehen-sive overview of diffusion-based collaborative filtering techniques, covering both 1) standard (non-sequential) recommendation and 2) sequential recommendation. We first explain the fundamental concepts of collaborative filtering and probabilistic diffusion models, and then summarize their applications to both standard and sequential recommendation settings.

关键词： Reviews Collaborative filtering Noise reduction Diffusion models Probabilistic logic Information filters Natural language processing Noise measurement Recommender systems Standards

来源：评论

学校读者我要写书评

暂无评论

Intrusion Detection Systems to avoid Cyberattack using Machine Learning 7

Intrusion Detection Systems to avoid Cyberattack using Machi...

引用

7th International Conference on Contemporary computing and Informatics, IC3I 2024

作者： Paithankar, Om Nandkumar Haider, Armaan Mohammed Suman, Preetam Padhy, Sasmita Siddiqui, Masood H Dash, Sachikanta Vit Bhopal University School of Computing Science and Engineering Madhya Pradesh Bhopal India University of Lucknow Department of Statistics Uttar Pradesh Lucknow India Giet University Department of Cse Odisha Gunupur India

ISBN: (纸本)9798350350067

Machine learning brings another perspective into intrusion detection systems. Conventional techniques stumble at the pace of fast-evolving cyber threats, so we delve into machine learning-based intrusion detection systems. In this paper, we survey different datasets using different models: supervised, unsupervised, and hybrid. The work presents feature selection, model optimization, and performance metrics. This paper handles well-known challenges connected with imbalanced data and risks of overfitting. Ensembling and deep learning realize stunning improvements over the traditional approaches. These findings offer in-depth views of practical ways in which machine learning can be integrated into network security with a clear way forward toward real adaptive resilient intrusion detection, underpinning the transformative role of ML in hardening defenses within the cyber domain against sophisticated attacks. © 2024 IEEE.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

The 1st ChatGPT4PCG Competition

引用

IEEE Transactions on Games 2024年 1-17页

作者： Abdullah, Febri Taveekitworachai, Pittawat Dewantoro, Mury F. Thawonmas, Ruck Togelius, Julian Renz, Jochen Graduate School of Information Science and Engineering Ritsumeikan University Kusatsu Japan College of Information Science and Engineering Ritsumeikan University Kusatsu Japan School of Computing The Australian National University Canberra Australia

This study summarizes the first ChatGPT4PCG competition held at the 2023 IEEE Conference on Games. The goal of the competition is to explore emergent abilities of publicly available LLMs in performing complex tasks related to procedural content generation, specifically physics-based level generation for Angry Bird-like games. Participants are tasked with submitting their prompts for ChatGPT to generate Angry Birds-like game structures that resemble English uppercase characters. A structure is a collection of stacked game objects comprising a part of an entire Angry Birds-like level. A prompt is an input for large language models (LLMs) including ChatGPT. Two evaluation metrics, i.e., stability and similarity, are used to evaluate the submitted prompts. Stability measures the sturdiness of a structure to withstand in-game gravity, while similarity measures a structure's resemblance to the target character. With such evaluation, participants are challenged not only to produce character-like but also stable structures by utilizing prompt engineering techniques. Finally, the competition's results are discussed to provide valuable insights for future studies and competitions. IEEE

关键词： Birds

来源：评论

学校读者我要写书评

暂无评论

Effective Integration of Text Diffusion and Pre-Trained Language Models with Linguistic Easy-First Schedule 30

Effective Integration of Text Diffusion and Pre-Trained Lang...

引用

Joint 30th International Conference on Computational Linguistics and 14th International Conference on Language Resources and Evaluation, LREC-COLING 2024

作者： Ou, Yimin Jian, Ping School of Computer Science and Technology Beijing Institute of Technology Beijing China Beijing Engineering Research Center of High Volume Language Information Processing and Cloud Computing Applications Beijing Institute of Technology Beijing China

ISBN: (纸本)9782493814104

Diffusion models have become a powerful generative modeling paradigm, achieving great success in continuous data patterns. However, the discrete nature of text data results in compatibility issues between continuous diffusion models (CDMs) and pre-trained language models (PLMs). That is, the performance of diffusion models even degrades when combined with PLMs. To alleviate this issue, we propose to utilize a pre-trained decoder to convert the denoised embedding vectors into natural language instead of using the widely used rounding operation. In this way, CDMs can be more effectively combined with PLMs. Additionally, considering that existing noise schedules in text diffusion models do not take into account the linguistic differences among tokens, which violates the easy-first policy for text generation, we propose a linguistic easy-first schedule that incorporates the measure of word importance, conforming to easy-first-generation linguistic features and bringing about improved generation quality. Experiment results on the E2E dataset and five controllable tasks show that our approach can combine the merits of CDMs and PLMs, significantly outperforming other diffusion-based models. © 2024 ELRA Language Resource Association: CC BY-NC 4.0.

关键词： Diffusion

来源：评论

学校读者我要写书评

暂无评论

Chicken Swarm Algorithm with Deep Learning for Plant Leaf Disease Detection and Classification 23th

Chicken Swarm Algorithm with Deep Learning for Plant Leaf Di...

引用

23rd International Conference on Intelligent Systems Design and Applications, ISDA 2023

作者： Subbiah, Priyanga Tyagi, Amit Kumar Krishnaraj, N. Department of Networking and Communications School of Computing Faculty of Engineering and Technology SRM Institute of Science and Technology Kattankulathur Chengalpattu603203 India Department of Fashion Technology National Institute of Fashion Technology New Delhi India

ISBN: (纸本)9783031648359

Diseases that affect plants have a major influence on crop yields around the globe. Crop yields and plant development have been severely limited in many parts of the world due to plant diseases for quite some time. The food supply has been negatively affected as a result of this. In certain cases, depending on the specifics of the issue, it may be possible to make out the plant’s leaves. Photo documentation of the plant’s leaves revealed the presence of many illnesses. It is crucial to determine the type of infestation before initiating any action to eradicate it. Among the many factors considered in this case study are the rapid spread of illnesses and the farmers’ lack of diagnostic expertise. Fortunately, by merging deep learning (DL) techniques into computer vision (CV), we might be able to simplify machine learning (ML) and so lessen this issue. The purpose of this study is to identify plant leaf diseases using a Deep Learning system in conjunction with a Chicken Swarm Algorithm. Helping farmers increase agricultural output while decreasing crop losses is one of the aims of the proposed method. Visual descriptions of images are used to achieve this. Accurately classifying the occurrence of diseases that commonly impact leaves is the goal of the proposed technique. The goal of this comprehensive experimental study was to compare the proposed method to a reference database and find out if it had any unique features. Based on the testing findings, the proposed algorithm outperformed the previous systems on multiple performance metrics. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

关键词： Diseases

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：