权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

SaTC: CORE: Small: Attack-Agnostic Defenses against Adversarial Inputs in Learning Systems

SaTC：核心：小：针对学习系统中的对抗性输入的与攻击无关的防御

基本信息

批准号：
1718787
负责人：
Ting Wang
金额：
$ 49.83万
依托单位：
Lehigh University
依托单位国家：
美国
项目类别：
Standard Grant
财政年份：
2017
资助国家：
美国
起止时间：
2017-08-15 至 2019-11-30
项目状态：
已结题

来源：
https://www.nsf.gov/awardsearch/showAward?AWD_ID=1718787&HistoricalAwards=false
关键词：
SaTC CORE Small Attack Agnostic

项目摘要

Deep learning technologies hold great promise to revolutionize the way people live and work. However, deep learning systems are inherently vulnerable to adversarial inputs, which are maliciously crafted samples to trigger deep neural networks to misbehave, leading to disastrous consequences in security-critical applications. The fundamental challenges of defending against such attacks stem from their adaptive and variable nature: adversarial inputs are tailored to target deep neural networks, while crafting strategies vary greatly with concrete attacks. This project develops EagleEye, a universal, attack-agnostic defense framework that (i) works effectively against unseen attack variants, (ii) preserves predictive power of deep neural networks, (iii) complements existing defense mechanisms, and (iv) provides comprehensive diagnosis about potential risks in deep learning outputs.In particular, EagleEye leverages a set of invariant properties underlying most attacks, including the "minimality principle": to maximize attack evasiveness, an adversarial input is generated by applying the minimum possible distortion to a legitimate input. By exploiting such properties in a principled manner, EagleEye effectively discriminates adversarial inputs (integrity checking) and even uncovers their correct outputs (truth recovery). The specific research tasks include: (i) identifying inherently distinct properties (differentiators) of legitimate and adversarial inputs, (ii) developing attack-agnostic adversarial input detection methods based on these differentiators, and (iii) analyzing possible countermeasures by adversaries to evade such defenses. This research not only facilitates the adoption of deep learning-powered systems and services, but also enlightens designing and implementing robust machine learning systems in general. New theories and systems developed in this project are integrated into undergraduate and graduate education and used to raise public awareness of the importance of machine learning security. More information about this project can be found at the project homepage: http://x-machine.github.io/project/eagleeye

深度学习技术有望彻底改变人们的生活和工作方式。然而，深度学习系统本质上容易受到对抗性输入的影响，这些输入是恶意制作的样本，可以触发深度神经网络的不当行为，从而在安全关键型应用中导致灾难性后果。防御此类攻击的根本挑战源于其自适应性和可变性：对抗性输入是针对深度神经网络量身定制的，而制定策略则因具体攻击而有很大差异。该项目开发了一个通用的、与攻击无关的防御框架，它（i）有效地对抗看不见的攻击变体，（ii）保留深度神经网络的预测能力，（iii）补充现有的防御机制，（iv）提供对深度学习输出中潜在风险的全面诊断。特别是，该框架利用了一组大多数攻击的不变属性，包括“最小原则”：为了最大化攻击规避，通过将最小可能失真应用于合法输入来生成对抗性输入。通过以有原则的方式利用这些属性，EagleEye有效地区分对抗性输入（完整性检查），甚至发现它们的正确输出（真相恢复）。具体的研究任务包括：（i）识别合法和对抗性输入的固有不同属性（区分器），（ii）基于这些区分器开发攻击不可知的对抗性输入检测方法，以及（iii）分析对手可能采取的对策以规避此类防御。这项研究不仅促进了深度学习驱动的系统和服务的采用，而且还启发了设计和实现健壮的机器学习系统。该项目开发的新理论和系统被整合到本科和研究生教育中，并用于提高公众对机器学习安全重要性的认识。有关此项目的更多信息，请访问项目主页：http://x-machine.github.io/project/eagleeye

项目成果

期刊论文数量（8）

专著数量（0）

科研奖励数量（0）

会议论文数量（0）

专利数量（0）

UNIFUZZ: A Holistic and Pragmatic Metrics-Driven Platform for Evaluating Fuzzers

DOI：
发表时间：
2020-10
期刊：
影响因子：
0
作者：
Yuwei Li;S. Ji;Yuan Chen;Sizhuang Liang;Wei-Han Lee;Yueyao Chen;Chenyang Lyu-;Chunming Wu;R. Be
通讯作者：
Yuwei Li;S. Ji;Yuan Chen;Sizhuang Liang;Wei-Han Lee;Yueyao Chen;Chenyang Lyu-;Chunming Wu;R. Be

DeepClean: Data Cleaning via Question Asking

DOI：
10.1109/dsaa.2018.00039
发表时间：
2018-10
期刊：
2018 IEEE 5th International Conference on Data Science and Advanced Analytics (DSAA)
影响因子：
0
作者：
Xinyang Zhang-;Yujie Ji;Chanh Nguyen;Ting Wang
通讯作者：
Xinyang Zhang-;Yujie Ji;Chanh Nguyen;Ting Wang

Integration of Static and Dynamic Code Stylometry Analysis for Programmer De-anonymization

集成静态和动态代码风格分析以实现程序员去匿名化

DOI：
10.1145/3270101.3270110
发表时间：
2018
期刊：
Proceedings of the 11th ACM Workshop on Artificial Intelligence and Security
影响因子：
0
作者：
Wang, Ningfei;Ji, Shouling;Wang, Ting
通讯作者：
Wang, Ting

Towards Evaluating the Security of Real-World Deployed Image CAPTCHAs

DOI：
10.1145/3270101.3270104
发表时间：
2018-01
期刊：
Proceedings of the 11th ACM Workshop on Artificial Intelligence and Security
影响因子：
0
作者：
Binbin Zhao;Haiqin Weng;S. Ji;Jianhai Chen;Ting Wang;Qinming He;Reheem Beyah
通讯作者：
Binbin Zhao;Haiqin Weng;S. Ji;Jianhai Chen;Ting Wang;Qinming He;Reheem Beyah

Towards Understanding the Dynamics of Adversarial Attacks

DOI：
10.1145/3243734.3278528
发表时间：
2018-10
期刊：
Proceedings of the 2018 ACM SIGSAC Conference on Computer and Communications Security
影响因子：
0
作者：
Yujie Ji;Ting Wang
通讯作者：
Yujie Ji;Ting Wang