서브메뉴
검색
Practical Application of Deep Generative Models to Network Traces and Use Cases
Practical Application of Deep Generative Models to Network Traces and Use Cases
Detailed Information
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211153024
- ISBN
- 9798342716444
- DDC
- 004
- 저자명
- Yin, Yucheng.
- 서명/저자
- Practical Application of Deep Generative Models to Network Traces and Use Cases
- 발행사항
- [Sl] : Carnegie Mellon University, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 151 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-05, Section: B.
- 주기사항
- Advisor: Sekar, Vyas.
- 학위논문주기
- Thesis (Ph.D.)--Carnegie Mellon University, 2024.
- 초록/해제
- 요약The recent advances of Deep Generative Models (DGMs) have emerged as a cornerstone of innovation and opened new frontiers across diverse domains. These models, which are designed to generate new data samples that closely resemble high-dimensional, real-world data using deep learning techniques, have been pivotal in their applications. Models like Generative Adversarial Networks (GANs), Diffusion Models, Transformers (e.g., GPT, BERT) have significantly contributed to generating high-quality, diverse samples in areas such as images, audio, and text data.Compared to images or text, an under-explored yet significant area where these models hold immense potential is the generation of network traces. Network traces encompass a wide range of data types, including packet/flow header traces, operational logs, security event logs etc. These datasets serve both near real-time management tasks for monitoring and troubleshooting (e.g., traffic engineering, anomaly detection) as well as more longitudinal tasks (e.g., traffic trends, forecasting, new pattern emergence).Building on these initial promise and observations, the overarching question we would like to ask in this thesis is: Can we practically apply these DGMs to generate realistic synthetic network traces and be potentially useful for various downstream tasks and use cases? Given the promising vision, there have been a series of practical challenges: Fidelity, Scalability, Privacy, Usability. In this thesis, as the first step towards the vision, we tackle the challenges above on various types of network traces and DGMs with different focuses. We make three key contributions: (1) We first introduce NetShare, a GAN-based synthetic IP header trace generator. (2) Second, we introduce DeepStore, a framework for evaluating the viability of deep generative compression for long-term network trace storage. (3) Third, we introduce CANGen, a domain-specific generator that focuses on in-vehicle CAN trace generation.
- 일반주제명
- Computer science
- 일반주제명
- Computer engineering
- 일반주제명
- Electrical engineering
- 키워드
- Machine learning
- 기타저자
- Carnegie Mellon University Electrical and Computer Engineering
- 기본자료저록
- Dissertations Abstracts International. 86-05B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017164621
■00520250211153024
■006m o d
■007cr#unu||||||||
■020 ▼a9798342716444
■035 ▼a(MiAaPQ)AAI31633001
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a004
■1001 ▼aYin, Yucheng.
■24510▼aPractical Application of Deep Generative Models to Network Traces and Use Cases
■260 ▼a[Sl]▼bCarnegie Mellon University▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a151 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-05, Section: B.
■500 ▼aAdvisor: Sekar, Vyas.
■5021 ▼aThesis (Ph.D.)--Carnegie Mellon University, 2024.
■520 ▼aThe recent advances of Deep Generative Models (DGMs) have emerged as a cornerstone of innovation and opened new frontiers across diverse domains. These models, which are designed to generate new data samples that closely resemble high-dimensional, real-world data using deep learning techniques, have been pivotal in their applications. Models like Generative Adversarial Networks (GANs), Diffusion Models, Transformers (e.g., GPT, BERT) have significantly contributed to generating high-quality, diverse samples in areas such as images, audio, and text data.Compared to images or text, an under-explored yet significant area where these models hold immense potential is the generation of network traces. Network traces encompass a wide range of data types, including packet/flow header traces, operational logs, security event logs etc. These datasets serve both near real-time management tasks for monitoring and troubleshooting (e.g., traffic engineering, anomaly detection) as well as more longitudinal tasks (e.g., traffic trends, forecasting, new pattern emergence).Building on these initial promise and observations, the overarching question we would like to ask in this thesis is: Can we practically apply these DGMs to generate realistic synthetic network traces and be potentially useful for various downstream tasks and use cases? Given the promising vision, there have been a series of practical challenges: Fidelity, Scalability, Privacy, Usability. In this thesis, as the first step towards the vision, we tackle the challenges above on various types of network traces and DGMs with different focuses. We make three key contributions: (1) We first introduce NetShare, a GAN-based synthetic IP header trace generator. (2) Second, we introduce DeepStore, a framework for evaluating the viability of deep generative compression for long-term network trace storage. (3) Third, we introduce CANGen, a domain-specific generator that focuses on in-vehicle CAN trace generation.
■590 ▼aSchool code: 0041.
■650 4▼aComputer science
■650 4▼aComputer engineering
■650 4▼aElectrical engineering
■653 ▼aDeep generative model
■653 ▼aMachine learning
■653 ▼aNetwork trace generation
■653 ▼aNetwork trace storage
■653 ▼aSynthetic data generation
■690 ▼a0984
■690 ▼a0544
■690 ▼a0464
■71020▼aCarnegie Mellon University▼bElectrical and Computer Engineering.
■7730 ▼tDissertations Abstracts International▼g86-05B.
■790 ▼a0041
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17164621▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
Preview
Export
ChatGPT Discussion
AI Recommended Related Books
Подробнее информация.
- Бронирование
- не существует
- моя папка
- Первый запрос зрения
- Non-Book Loan Application
- Nighttime Book Loan Application
Available after logging in.


