Virtual Triggering a Technique to Segment Cryptographic Processes in Side-Channel Traces 1stJeremy Guillaume

2025-05-06 0 0 3.11MB 6 页 10玖币

侵权投诉

Virtual Triggering: a Technique to Segment

Cryptographic Processes in Side-Channel Traces

1st Jeremy Guillaume

IETR UMR CNRS 6164

CentraleSup´

elec Rennes Campus

35576 Cesson-Sevign´

e, France

jeremy.guillaume@centralesupelec.fr

2nd Maxime Pelcat

IETR UMR CNRS 6164

INSA Rennes

35700 Rennes, France

maxime.pelcat@insa-rennes.fr

3rd Amor Nafkha

IETR UMR CNRS 6164

CentraleSup´

elec Rennes Campus

35576 Cesson-Sevign´

e, France

amor.nafkha@centralesupelec.fr

4th Rub´

en Salvador

IETR UMR CNRS 6164

CentraleSup´

elec Rennes Campus

35576 Cesson-Sevign´

e, France

ruben.salvador@centralesupelec.fr

Abstract—Side-Channel Attacks (SCAs) exploit data correla-

tion in signals leaked from devices to jeopardize conﬁdentiality.

Locating and synchronizing segments of interest in traces from

Cryptographic Processes (CPs) is a key step of the attack. The

most common method consists in generating a trigger signal

to indicate to the attacker the start of a CP. This paper

proposes a method called Virtual Triggering (VT) that removes

the need for the trigger signal and automates trace segmentation.

When the time between repetitions is not constant, further trace

alignment techniques are required. Building on VT, we propose

a simple method to learn representative segment templates from

a proﬁling device similar to the victim, and to automatically

locate and pull out these segments from other victim devices

using simple pattern recognition. We evaluate VT on screaming

channel attacks [1], which initially used a Frequency Component

(FC) known to appear at a single time in leaked signals, as a

trigger to segment traces. We demonstrate that VT not only

performs equivalently to FC on a standard attack scenario, but

we also show how using VT with the automatic pullout technique

improves the attack efﬁciency and enables more realistic attack

scenarios. Thanks to VT, screaming channel attacks can now:

(1) succeed with only half of the segments collected compared

to the FC trigger from the original attack; and (2) absorb time

variations between CPs.

Index Terms—Cybersecurity, side-channel attacks, screaming

channels, electromagnetic side-channels, trace collection.

I. INTRODUCTION

Side-Channel Attacks (SCAs) [2] exploit data-correlated

leakages that arise when a device operates on data. The

term side-channel is used to denote physical leakage signals

carrying conﬁdential information. Side-channels are inherent

to CMOS computing devices and can take many forms,

from timing to power consumption to Electromagnetic (EM)

emanations. We call traces to the sampled measurements of a

side-channel during the execution of one or multiple targeted

operations. The most common scenario in SCAs targets a

cryptographic key manipulated by Cryptographic Processes

(CPs), as eavesdropping such a key can jeopardize system

conﬁdentiality. We refer to these attacks as Side-Channel

Cryptographic Attacks (SCCAs) [3]. A trace can contain

leakage from multiple CPs. Trace segmentation separates the

different segments from the leakage, each corresponding to

the side-channel measurement of one CP execution. During

the attacking phase, a hypothesis ˆy∈Yon the data yis made

(Y= all possible data values). Techniques like Differential

Power Analysis (DPA) [4], Correlation Power Analysis (CPA)

[5], Mutual Information Analysis (MIA) [6], template attacks

[7] and more recently Deep Learning (DL) [8], evaluate the

relationship between ˆyand the segment leakage values. Each

hypothesis Yis tested, and a probability is returned for all

of them. The ˆyhaving the highest probability is expected

to correspond to the data y. For these attacks to work, it

is important to know, for each segment point, to which CP

operations they belong. It makes it possible to ﬁnd a relation

between yand leak values of the same CP data-correlated

operations present in all segments. To respect this requirement,

segment synchronization is done during the collecting and pre-

processing phases of the attack.

Focused on screaming channel attacks, we look into its

collecting and pre-preprocessing phases, which aim at obtain-

ing synchronized and denoised segments from the victim’s

leakage traces using techniques like time diversity (average

multiple CPs computing the same data to reduce the noise).

To relax the triggering requirements for trace capturing and

synchronization, our contributions include:

•Virtual Triggering (VT), a method to segment traces from

side-channels of an embedded device executing a cryp-

tographic software implementation. VT requires neither

external synchronization nor tampered victim software.

•an experimental evaluation of the proposal on a realistic

arXiv:2210.12059v1 [cs.CR] 21 Oct 2022

Analogue

Substrate

Legitimate signal + Noise

Substrate Coupling

DAC Amp

Digital Noise Fclock FRadio Fclock

Fclock

Digital

Fig. 1. Screaming channel attacks: Conventional side-channels leak to the

RF module in the analog part, present on the same die. This one transmits

side-channels at a larger distance (until some meters).

dard (AES) on an embedded device.

•a discussion on the method limitations and solutions.

•experimental results demonstrating the gains obtained

with these solutions.

VT does not require any speciﬁc setup on the victim side,

like a trigger signal to indicate the start of a CP. It consists in

ﬁnding a precise enough time duration of the targeted process

executed periodically. This makes it possible to act as if a

trigger would indicate a common location in all the process

segments. This virtual trigger can then be used to segment

CPs from a trace. The method aims at helping researchers to

reduce the effort in target preparation and in the collecting

phase of the attack, while also giving a small step toward a

more realistic attack scenario.

The paper is organized as follows. Section II provides the

context of this work and Section III discusses related works.

Then, Section V details the proposed virtual trigger segmenta-

tion method, and Section VI evaluates it experimentally on a

screaming channel setup. Section VII discusses the limitation

of the method and proposes a solution to overcome it. Finally,

Section VIII concludes the paper.

II. SCREAMING CHANNEL ATTACKS

Experimental results build on the attack scenario called

screaming channels introduced by Camurati, et al. [1]. As

illustrated in Fig. 1, screaming channels occur on mixed-signal

devices where digital processing is collocated with analog

Radio Frequency (RF) electronics over a single die. Side-

channels originated from digital processing mix with RF signal

and get ampliﬁed, modulated, and broadcast. The primary

threat posed by screaming channels is the risk of transmitting

secrets over long distances, i.e., scream them.

The screaming channel signals are very noisy. Plus, to

collect them, it is necessary that the RF module transmits a

legitimate signal. In the context of this paper, it is a Bluetooth

signal. Between two Bluetooth transmissions, the collected

signal contains holes. As in regular screaming channel analy-

sis, to counterbalance these constraints, time diversity is used

during the collection phase. The principle is to force the device

to compute multiple encryptions with the same plaintext and

key. Since the same data has been computed, their segment

values should be the same, except for the noise. Averaging

the segments returns a CP segment with reduced noise.

III. RELATED WORKS ON SEGMENT SYNCHRONIZATION

Synchronization is used to know to which CP operations

each segment point belongs to. Otherwise, segment points

corresponding to operations whose leakage values are data-

correlated would be compared with other unrelated points.

Therefore, it would be harder to distinguish a relationship be-

tween leakage values and data. We name these data-correlated

points as Points of Interest (POIs). To synchronize segments,

an alignment between them can be done using techniques like

static alignment [9], longest common sequence [10], elastic

alignment [11] and synchronous real-time sampling [12].

Before aligning segments, it is ﬁrst necessary to locate these

segments of interest in the traces. The most common technique

in SCA research consists of inserting a trigger signal to start

the trace measurement synchronized with the beginning of

the CP. Attack setups are prepared to either have (1) the

victim to create the trigger signal to inform the attacker when

encryption starts, or (2) the attacker sending a trigger signal

to the victim to make it start at a precise moment. This is an

accepted scenario in the community to enable SCA research.

But it assumes attackers have access to the victim to generate

or listen to a trigger synchronization signal. SAKURA and

NewAE’s Chipwhisperer are widely used platforms in the

community that follow this approach.

However, in many cases, using a trigger signal is impossible.

For example, when a given ﬁrmware cannot be modiﬁed to

add instructions that control the trigger signal. Or simply

because the device used to collect traces is not capable of

capturing two signals, the side-channel signal and the trigger

signal, concurrently. Locating CPs in traces without using

trigger signals can be done with pattern recognition techniques.

For this purpose, Beckers, et al. [13] compare methods that

calculate the correspondence between trace and pattern values.

When this match score is over a pre-deﬁned threshold, the

corresponding part of the leakage is considered as being

the location in the trace of a targeted segment. IcWaves1

implements such pattern recognition methods.

Nevertheless, to use these methods, the attacker is supposed

to already have a pattern or characterized segments having

the same statistical properties as the researched segments,

representative of the triggering moment. Therefore, the ques-

tion of how to ﬁnd this pattern remains open. To that end,

Trautmann et al. [14] proposed a technique to locate AES

CPs in leakage signals by searching for parts of the leakage

having consecutive similar patterns corresponding to the 10

AES rounds. This method can ﬁnd AES CPs in long traces

also containing other CP operation leakages. Souissi et al. [15]

used wavelet transforms to detect the limit of AES segments in

traces and then used these segments to do pattern recognition.

In the screaming channels context, in order to locate CPs

from leakage signals, the only technique reported so far in

the literature, by Camurati et al. [1] and Wang et al. [16],

used a frequency component trigger mechanism2. The method

1https://www.riscure.com/security-tools/hardware/icwaves.

2https://github.com/bolek42/rsa-sdr.

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：

VirtualTriggering:aTechniquetoSegmentCryptographicProcessesinSide-ChannelTraces1stJeremyGuillaumeIETRUMRCNRS6164CentraleSup´elecRennesCampus35576Cesson-Sevign´e,Francejeremy.guillaume@centralesupelec.fr2ndMaximePelcatIETRUMRCNRS6164INSARennes35700Rennes,Francemaxime.pelcat@insa-rennes.fr3rdAmorNafkh...

展开>> 收起<<

Virtual Triggering a Technique to Segment Cryptographic Processes in Side-Channel Traces 1stJeremy Guillaume.pdf

共6页,预览2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！

Virtual Triggering a Technique to Segment Cryptographic Processes in Side-Channel Traces 1stJeremy Guillaume

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: