Title

Microarchitectural Vulnerabilities of Performance and Energy Efficiency Techniques in Data Center Servers

Alternative Title

데이터 센터 서버 내 성능 및 에너지 효율 기술의 마이크로아키텍처 취약점

DGIST Authors

Hyosang Kim ; Daewon Seo ; Donghoon Shin

Advisor

서대원

Co-Advisor(s)

Donghoon Shin

Issued Date

2026

Awarded Date

2026-08-01

Type

Thesis

Description

Microarchitectural Security, Idle states, Accelerator, Covert Channel, Side-channel|마이크로아키텍처 보안, 유휴 상태, 가속기, 은닉 채널, 부채널

Table Of Contents

I. Introduction 1
1.1 Contributions 2
1.2 Organization 3
II. Background 4
2.1 Idle State Management 4
2.2 Network Packet Processing 5
2.3 Intel Data Streaming Accelerator 6
2.4 Covert and Side-channels 7
III. Related Work 8
3.1 Cache-Based Channels 8
3.2 Power-Based Channels 9
3.3 Memory, GPU, and FPGA-Based Channels 11
3.4 On-chip Accelerator-Based Channels 11
IV. BrokenSleep: Remote Power Timing Attack Exploiting Processor Idle States 12
4.1 C-states Analysis 12
4.1.1 Wake-up Latency of Core C-states 12
4.1.2 Impact of Core C-states on Request Processing 13
4.2 Threat Model 15
4.3 Covert Channel 16
4.3.1 Mechanism 16
4.3.2 Methodology 18
4.3.3 Evaluation Results 20
4.4 Side-channel: Inferring Keystroke Activities 23
4.4.1 Mechanism 23
4.4.2 Methodology 24
4.4.3 Evaluation Results 25
4.5 Mitigation Strategies for C-state vulnerabilities 26
4.6 Summary 27
V. DarkStream: Exploiting Internal Throughput Contention in Data Streaming Accelerator for Timing Attacks 29
5.1 Internal Contention in DSA 29
5.2 Covert Channel exploiting Intel DSA 32
5.2.1 Threat Model 32
5.2.2 Characterization of Timing Variability Across Operations 33
5.2.3 Analysis of Memory Move as a Covert Channel Primitive 34
5.2.4 DarkStream Covert Channel 35
5.2.5 Methodology 36
5.2.6 Experimental Results 37
5.3 Side-channel 41
5.3.1 Threat Model 42
5.3.2 Approach 42
5.3.3 Website Fingerprinting 47
5.3.4 Deep Learning Model Fingerprinting 49
5.4 Countermeasures 50
5.5 Summary 51
VI. Conclusion 53
References 55
요약문 62

URI

https://scholar.dgist.ac.kr/handle/20.500.11750/60748
http://dgist.dcollection.net/common/orgView/200001006857

DOI

10.22677/THESIS.200001006857

Degree

Doctor

Department

Department of Electrical Engineering and Computer Science

Publisher

DGIST

Show Full Item Record