SpecEES: Scaling Deep Learning Models for Cellular Spectrum Monitoring

SpecEES：扩展用于蜂窝频谱监控的深度学习模型

• 批准号: 1923778
• 负责人: Haitao Zheng
• 金额: $ 75万
• 依托单位: University of Chicago
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-09-01 至 2023-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1923778&HistoricalAwards=false
• 关键词: SpecEES; Scaling; Deep; Learning; Models

Spectrum management in wireless networks is a challenging task that will only increase in difficulty as complexity grows in hardware, configurations, and new access technology. Government agencies and wireless providers need robust and flexible tools to monitor and detect anomalies (i.e., both faults and misbehavior) in physical spectrum usage, and to deploy them at scale. This project targets the open challenge of spectrum anomaly detection for wide-area cellular networks, providing a practical architecture to monitor, diagnose, and secure spectrum usage. The project will significantly improve spectrum efficiency, while tackling multiple open challenges related to energy efficiency and security. The technical component of the project will be tightly integrated with educational and outreach programs to engage female and underrepresented students into research, while offering a mentoring platform. The project will develop and strengthen collaborations with academic, government and industry partners, across multiple areas of spectrum measurements, energy-efficient systems, security, and applied machine learning.The core concept driving this project is the integration of a distributed spectrum measurement platform with an efficient deep neural network-based anomaly detection framework. Deep learning in this context introduces multiple challenges, including scalability, lack of location-specific training data, lack of misuse events in data, and potential for adversarial attacks. Our proposed work addresses these with a combination of context-agnostic training, transfer learning, semi-supervised clustering for detection of anomalies and adversarial countermeasures. The proposed system will incorporate spectrum measurements taken by various static and mobile observers, and use these measurements to train an efficient, scalable, and robust anomaly detection module driven by deep neural network models. The detection module will run on both static and mobile observers, allowing the system to detect and diagnose spectrum anomalies in real-time.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

无线网络中的频谱管理是一项具有挑战性的任务，随着硬件、配置和新接入技术的复杂性增加，其难度只会增加。政府机构和无线供应商需要强大而灵活的工具来监控和检测物理频谱使用中的异常情况(即故障和不当行为)，并大规模部署它们。该项目针对广域蜂窝网络频谱异常检测的开放挑战，提供了一个实用的架构来监控、诊断和保护频谱使用。该项目将显著提高频谱效率，同时应对与能源效率和安全相关的多重开放挑战。该项目的技术部分将与教育和推广计划紧密结合，以吸引女性和代表性不足的学生参与研究，同时提供一个指导平台。该项目将发展和加强与学术、政府和行业合作伙伴的合作，涉及频谱测量、节能系统、安全和应用机器学习等多个领域。推动该项目的核心概念是将分布式频谱测量平台与高效的基于深度神经网络的异常检测框架相集成。在这种情况下的深度学习带来了多重挑战，包括可伸缩性、缺乏特定位置的训练数据、缺乏数据中的误用事件以及潜在的对抗性攻击。我们提出的工作结合了上下文不可知训练、迁移学习、用于检测异常的半监督聚类和对抗性对策来解决这些问题。该系统将结合各种静态和移动观察者的频谱测量，并使用这些测量来训练一个由深度神经网络模型驱动的高效、可扩展和健壮的异常检测模块。检测模块将在静态和移动观察者上运行，允许系统实时检测和诊断频谱异常。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

Understanding Robust Learning through the Lens of Representation Similarities

• DOI: 10.48550/arxiv.2206.09868
• 发表时间: 2022-06
• 期刊: ArXiv
• 作者: Christian Cianfarani;A. Bhagoji;Vikash Sehwag;Ben Y. Zhao;Prateek Mittal

Understanding the Effect of Bias in Deep Anomaly Detection

• DOI: 10.24963/ijcai.2021/456
• 发表时间: 2021-05
• 期刊: ArXiv
• 作者: Ziyu Ye;Yuxin Chen;Haitao Zheng

Post-breach Recovery: Protection against White-box Adversarial Examples for Leaked DNN Models

• DOI: 10.1145/3548606.3560561
• 发表时间: 2022-05
• 期刊: Proceedings of the 2022 ACM SIGSAC Conference on Computer and Communications Security
• 作者: Shawn Shan;Wen-Luan Ding;Emily Wenger;Haitao Zheng;Ben Y. Zhao

Gotta Catch'Em All: Using Honeypots to Catch Adversarial Attacks on Neural Networks

• DOI: 10.1145/3372297.3417231
• 发表时间: 2019-04
• 期刊: Proceedings of the 2020 ACM SIGSAC Conference on Computer and Communications Security
• 作者: Shawn Shan;Emily Wenger;Bolun Wang;B. Li;Haitao Zheng;Ben Y. Zhao

Fawkes: Protecting Privacy against Unauthorized Deep Learning Models

• 发表时间: 2020-02
• 作者: Shawn Shan;Emily Wenger;Jiayun Zhang;Huiying Li;Haitao Zheng;Ben Y. Zhao