adversarial backdoored purifies neurips
NeurIPS 2023 | 腾讯 AI Lab 18 篇入选论文解读
前言 NeurIPS 2023(Neural Information Processing Systems)神经信息处理系统大会是当前全球最负盛名的 AI 学术会议之一,将于 12 月 10 日在美国新奥尔良召开。官方信息显示,本届会议共有 12343 篇有效论文投稿,接收率为 26.1%,略高于 ......
【五期杨志】CCF-A(CVPR'22) Dual-Key Multimodal Backdoors for Visual Question Answering
Walmer M, Sikka K, Sur I, et al. Dual-Key Multimodal Backdoors for Visual Question Answering[C]//Proceedings of the IEEE/CVF Conference on Computer Vi ......
论文阅读-Self-supervised and Interpretable Data Cleaning with Sequence Generative Adversarial Networks
1. GARF 简介 代码地址:https://github.com/PJinfeng/Garf-master 基于 SeqGAN 提出了一种自监督、数据驱动的数据清洗框架——GARF。 GARF 的数据清洗分为两个步骤: 规则生成 (Rule generation with SeqGAN):利用 ......
sans sec 565 Red Team Operations and Adversary Emulation - 红队运营和对手仿真 之 565.1 Lab 1.4:奖金!用户名枚举和密码喷射
565.1 Lab 1.4:用户名枚举和密码喷射 目标 用户名枚举以发现其他有效用户 使用已知密码对新发现的账户进行喷洒 本实验室模拟的 TTP T1594 - Search Victim-Owned Websites T1078 - Valid Accounts T1087.003 - Accou ......
sans sec 565 Red Team Operations and Adversary Emulation - 红队运营和对手仿真 之 565.1 Lab 1.3:侦察和密码攻击
sans sec 565 Red Team Operations and Adversary Emulation - 红队运营和对手仿真 之 565.1 Lab 1.3:侦察和密码攻击 目标 通过分析 Draconem.io 网站进行侦察 确定密码攻击的目标对象 通过收集电子邮件地址发现有效的用户名 ......
sans sec 564 Red Team Operations and Adversary Emulation - 红队运营和对手仿真
564.1 红队演习介绍与规划 混乱的术语定义: 不需要知道这些词语的分别含义,只需要知道你在搞渗透 • Ethical Hacking • Vulnerability Scanning • Vulnerability Assessment(SEC460: Enterprise Threat and ......
NeurIPS 2023 | 清华ETH提出首个二值化光谱重建算法
前言 本文首次探索了压缩量化在光谱压缩重建领域的应用,提出了该领域首个二值化卷积神经网络 BiSRNet,在量化指标和视觉结果上都显著地超越了当前最先进的二值化模型。 本文转载自我爱计算机视觉 仅用于学术分享,若侵权请联系删除 欢迎关注公众号CV技术指南,专注于计算机视觉的技术总结、最新技术跟踪、经 ......
study of 'Missing data imputation framework for bridge structural health monitoring based on slim generative adversarial networks'
the Stochastic Gradient Descent (SGD):为了提高鲁棒性,SGAIN框架的优化器采用了随机梯度下降(SGD) 一,SGAIN框架有两个重要目的:鉴别器D的目的是最大化正确预测M矩阵的概率;生成器的目的是最小化D预测M矩阵的概率。此外,利用反向传播算法对发生器和鉴别器 ......
NeurIPS 2023 Spotlight | 半监督与扩散模型结合,实现少标签下可控生成
前言 本文从模型结构到训练策略,以及实验结果出发,详解了DeepMind之前提出的不需要归一化的深度学习模型NFNet。 本文转载自PaperWeekly 作者:游泽彬 单位:中国人民大学 仅用于学术分享,若侵权请联系删除 欢迎关注公众号CV技术指南,专注于计算机视觉的技术总结、最新技术跟踪、经典论 ......
NeurIPS 2023 | 「解释一切」图像概念解释器来了,港科大团队出品
前言 Segment Anything Model(SAM)首次被应用到了基于增强概念的可解释 AI 上。 本文转载自机器之心 仅用于学术分享,若侵权请联系删除 欢迎关注公众号CV技术指南,专注于计算机视觉的技术总结、最新技术跟踪、经典论文解读、CV招聘信息。 CV各大方向专栏与各个部署框架最全教程 ......
GAN(生成对抗网络,Generative Adversarial Network)
生成对抗网络(GAN)是一种深度学习模型架构,由生成器(Generator)和判别器(Discriminator)两个神经网络组成。这两个网络之间进行博弈式训练。 生成器(Generator):生成器是一个神经网络模型,它接收一个随机噪声向量作为输入,并试图生成与训练数据相似的新数据样本。生成器的目 ......
Proj CDeepFuzz Paper Reading: PELICAN: Exploiting Backdoors of Naturally Trained Deep Learning Models In Binary Code Analysis
## Abstract 背景: 1. 本文研究的不是被恶意植入的后门,而是products of defects in training 2. 攻击模式: injecting some small fixed input pattern(backdoor) to induce misclassifi ......
【五期邹昱夫】CCF-A(TIFS'23)SAFELearning: Secure Aggregation in Federated Learning with Backdoor Detectability
> "Zhang, Zhuosheng, et al. "SAFELearning: Secure Aggregation in Federated Learning with Backdoor Detectability." IEEE Transactions on Information For ......
论文解读(AdSPT)《Adversarial Soft Prompt Tuning for Cross-Domain Sentiment Analysis》
Note:[ wechat:Y466551 | 可加勿骚扰,付费咨询 ] 论文信息 论文标题:Adversarial Soft Prompt Tuning for Cross-Domain Sentiment Analysis论文作者:Hui Wu、Xiaodong Shi论文来源:2022 ACL ......
【五期邹昱夫】CCF-A(SP'23)3DFed: Adaptive and Extensible Framework for Covert Backdoor Attack in Federated Learning
> "Li, Haoyang, et al. "3DFed: Adaptive and Extensible Framework for Covert Backdoor Attack in Federated Learning." 2023 IEEE Symposium on Security an ......
论文解读(MCADA)《Multicomponent Adversarial Domain Adaptation: A General Framework》
Note:[ wechat:Y466551 | 可加勿骚扰,付费咨询 ] 论文信息 论文标题:Multicomponent Adversarial Domain Adaptation: A General Framework论文作者:Chang’an Yi, Haotian Chen, Yonghu ......
论文解读(TAT)《 Transferable Adversarial Training: A General Approach to Adapting Deep Classifiers》
Note:[ wechat:Y466551 | 可加勿骚扰,付费咨询 ] 论文信息 论文标题:Transferable Adversarial Training: A General Approach to Adapting Deep Classifiers论文作者:Hong Liu, Mingsh ......
论文解读(Moka‑ADA)《Moka‑ADA: adversarial domain adaptation with model‑oriented knowledge adaptation for cross‑domain sentiment analysis》
Note:[ wechat:Y466551 | 可加勿骚扰,付费咨询 ] 论文信息 论文标题:Moka‑ADA: adversarial domain adaptation with model‑oriented knowledge adaptation for cross‑domain senti ......
《Universal and Transferable Adversarial Attacks on Aligned Language Models》论文学习
一、Abstract 尽管“开箱即用”的大型语言模型(例如ChatGPT)能够生成出色的处理令人反感的内容,人们在规避针对LLM的攻击(针对LLM的所谓“越狱”)方面取得了一些成功,但在不断地攻防实践中这些防御手段却很脆弱,研究员在自动对抗性提示(prompt)生成方面也取得了一些突破。 在本文中, ......
论文解读(BERT-DAAT)《Adversarial and Domain-Aware BERT for Cross-Domain Sentiment Analysis》
论文信息 论文标题:Adversarial and Domain-Aware BERT for Cross-Domain Sentiment Analysis论文作者:论文来源:2020 ACL论文地址:download 论文代码:download视屏讲解:click 1 介绍 2 问题定义 在跨域 ......
Adversarial Attack(对手的攻击)
Adversarial Attack(对手的攻击) 把训练好的神经网络用在应用上,还需要让其输入人为的恶意行为,要在有人试图欺骗他的情况下得到高的正确率 例如:影像辨识,输入的图片加入一些杂讯(这些杂讯可能肉眼看不出来),使得输出错误,并输入某个指定的错误输出 无目标攻击:使输出结果与正确答案的差距 ......
Wild Patterns: Ten Years After the Rise of Adversarial Machine Learning---reading
# Wild Patterns: Ten Years After the Rise of Adversarial Machine Learning reading - 攻击目标 - 安全破坏 - 完整性破坏: 逃避检测,而不影响正常的系统运行 - 可用性破坏: 使得合法用户不能正常使用系统 - 隐私 ......
【五期邹昱夫】CCF-B(IEEE Access'19)Badnets: Evaluating backdooring attacks on deep neural networks
> "Gu, Tianyu, et al. "Badnets: Evaluating backdooring attacks on deep neural networks." IEEE Access 7 (2019): 47230-47244." 本文提出了外包机器学习时选择值得信赖的提供商的重要 ......
【五期邹昱夫】CCF-B(RAID'18)Fine-Pruning: Defending Against Backdooring Attacks on Deep Neural Networks
> "Liu, Kang, Brendan Dolan-Gavitt, and Siddharth Garg. "Fine-pruning: Defending against backdooring attacks on deep neural networks." Research in Att ......
【五期邹昱夫】CCF-A(NeurIPS'22)Trap and Replace: Defending Backdoor Attacks by Trapping Them into an Easy-to-Replace Subnetwork
> "Wang, Haotao, et al. "Trap and Replace: Defending Backdoor Attacks by Trapping Them into an Easy-to-Replace Subnetwork." Advances in Neural Informa ......
【五期邹昱夫】CCF-A(NeurIPS'21)Adversarial Neuron Pruning Purifies Backdoored Deep Models
> "Wu, Dongxian, and Yisen Wang. "Adversarial neuron pruning purifies backdoored deep models." Advances in Neural Information Processing Systems 34 (2 ......
SNN-RAT: Robustness-enhanced Spiking Neural Network through Regularized Adversarial Training
郑重声明:原文参见标题,如有侵权,请联系作者,将会撤销发布! 同大组工作 Abstract ......
Robust Deep Reinforcement Learning through Adversarial Loss
郑重声明:原文参见标题,如有侵权,请联系作者,将会撤销发布! 35th Conference on Neural Information Processing Systems (NeurIPS 2021) Abstract 最近的研究表明,深度强化学习智能体很容易受到智能体输入上的小对抗性扰动的影响 ......
【五期邹昱夫】CCF-A(NeurIPS'19)Inverting gradients-how easy is it to break privacy in federated learning?
"Geiping J, Bauermeister H, Dröge H, et al. Inverting gradients-how easy is it to break privacy in federated learning?[J]. Advances in Neural Informat ......
[Pix2Pix] Image-to-Image Translation with Conditional Adversarial NetWorks
paper:https://arxiv.org/pdf/1611.07004.pdf [CVPR 2017] code: https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix https://phillipi.github.io/pix2pi ......