Short Paper Static and Microarchitectural ML-Based Approaches For Detecting SpectreVulnerabilities and At_tacks Chidera Biringa

2025-05-03 0 0 2.67MB 5 页 10玖币

侵权投诉

Short Paper: Static and Microarchitectural ML-Based Approaches

For Detecting Spectre Vulnerabilities and Aacks

Chidera Biringa

cbiringa@umassd.edu

University of Massachusetts

Dartmouth, USA

Gaspard Baye

bgaspard@umassd.edu

University of Massachusetts

Dartmouth, USA

Gökhan Kul

gkul@umassd.edu

University of Massachusetts

Dartmouth, USA

ABSTRACT

Spectre intrusions exploit speculative execution design vulnerabil-

ities in modern processors. The attacks violate the principles of

isolation in programs to gain unauthorized private user informa-

tion. Current state-of-the-art detection techniques utilize micro-

architectural features or vulnerable speculative code to detect these

threats. However, these techniques are insuﬃcient as Spectre at-

tacks have proven to be more stealthy with recently discovered

variants that bypass current mitigation mechanisms. Side-channels

generate distinct patterns in processor cache, and sensitive infor-

mation leakage is dependent on source code vulnerable to Spectre

attacks, where an adversary uses these vulnerabilities, such as

branch prediction, which causes a data breach. Previous studies

predominantly approach the detection of Spectre attacks using the

microarchitectural analysis, a reactive approach. Hence, in this pa-

per, we present the ﬁrst comprehensive evaluation of static and

microarchitectural analysis-assisted machine learning approaches

to detect Spectre vulnerable code snippets (preventive) and Spectre

attacks (reactive). We evaluate the performance trade-oﬀs in em-

ploying classiﬁers for detecting Spectre vulnerabilities and attacks.

CCS CONCEPTS

•Security and privacy

→

Static Code and Microarchitectural

Analysis; • Detection →Machine and Deep Learning.

KEYWORDS

Spectre Vulnerability, Spectre Attack, Gadgets, CPU Processes State

ACM Reference Format:

Chidera Biringa, Gaspard Baye, and Gökhan Kul. 2022. Short Paper: Static

and Microarchitectural ML-Based Approaches For Detecting Spectre Vul-

nerabilities and Attacks. In .,5pages.

1 INTRODUCTION

Speculative execution [

] is a

𝜇

arch method used to improve mod-

ern microprocessor performance. In 2018, Kocher et al. [

] showed

that components that support speculative execution of assembly in-

structions such as branch predictions leave quantiﬁable side eﬀects

in processor caches along with other shared resources even with

the absence of instruction commit. Spectre attacks [

] are a class of

𝜇

arch attacks that pose a signiﬁcant threat to a computer’s security

Permission to make digital or hard copies of all or part of this work for personal or

classroom use is granted without fee provided that copies are not made or distributed

for proﬁt or commercial advantage and that copies bear this notice and the full citation

on the ﬁrst page. Copyrights for components of this work owned by others than ACM

must be honored. Abstracting with credit is permitted. To copy otherwise, or republish,

to post on servers or to redistribute to lists, requires prior speciﬁc permission and/or a

fee. Request permissions from permissions@acm.org.

, ,

by revealing private user data through a side-channel cache-timing

attack. Spectre-variant attacks exploit processor branch prediction

to obtain the victim’s data. A holistic and aggressive ﬁx implies re-

thinking the contract between the instruction set architecture (ISA)

and

𝜇

arch [

]. This realization led to the development of several

detection [

] and mitigation [

] solutions. Currently,

𝜇

arch-based detection techniques are dependent on hardware per-

formance measures [

], which dictate the distribution of

cache stress levels – hits or misses over time. However, attacks can

still propagate with adversarial manipulation of the performance

counters proﬁling tools such as

perf

. On the other hand, detection

using Spectre vulnerable code snippets [

] is constrained to

known vulnerabilities and side-channel data breaches. Several Spec-

tre mitigation strategies such as LFENCE [

] and Kernel Page Table

Isolation (KPTI) [

] for solving Meltdown attacks [

] attempt to

obtain a viable concession between performance and security. This

approach institutes a strictly enforced security that invariably leads

to a sub-optimal processor performance [

]. Meltdown is an

attack variant similar in principle to Spectre that exploits pipelined

access to memory during out-of-order execution to compromise

and leak user data from the kernel mode of the CPU. Recent studies

in the literature [

] have approached the problem of

Spectre attacks as a learning problem with malicious and benign

classes and consequently applying Machine Learning (ML) classi-

ﬁers to detect this attacks using either vulnerable victim programs

or hardware performance counters (HPCs).

In this study, we are motivated by: (i) the recent availability of

a signiﬁcant volume of Spectre gadgets to perform Spectre vul-

nerabilities detection using ML [

], and (ii) a gap in a thorough

presentation of the trade-oﬀs in performance between ML classi-

ﬁers using vulnerabilities and attacks data. Before 2021, conducting

ML-assisted Spectre vulnerabilities experiments via victim program

was constrained to no more than 17 observations [

], which is

not ideal and attributed to the fact that ML classiﬁers, especially

neural networks require large volumes of data to be suﬃciently

trained and explored [

]. Tol et al. [

] solved this problem by using

a combination of mutational fuzzing and deep learning (DL) to gen-

erate a signiﬁcant number of Spectre-V1 gadgets suitable for ML

and DL experiments. We approach detecting Spectre attacks from

both signature vulnerabilities in code and behavioral characteristics

in CPU-Processes State (CPS) by leveraging the traces of malicious

activity caused by Spectre attacks in the

𝜇

arch and detecting Spectre

vulnerabilities using gadgets. We propose a comprehensive perfor-

mance evaluation of static and cache analysis-assisted machine

learning approaches to detect Spectre-vulnerable programs and

attacks.

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

ShortPaper:StaticandMicroarchitecturalML-BasedApproachesForDetectingSpectreVulnerabilitiesandAacksChideraBiringacbiringa@umassd.eduUniversityofMassachusettsDartmouth,USAGaspardBayebgaspard@umassd.eduUniversityofMassachusettsDartmouth,USAGökhanKulgkul@umassd.eduUniversityofMassachusettsDartmouth,USA...

展开>> 收起<<

Short Paper Static and Microarchitectural ML-Based Approaches For Detecting SpectreVulnerabilities and At_tacks Chidera Biringa.pdf

共5页,预览1页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Short Paper Static and Microarchitectural ML-Based Approaches For Detecting SpectreVulnerabilities and At_tacks Chidera Biringa

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: