A General Security Approach for Soft-information Decoding against Smart Bursty Jammers Furkan Ercany Kevin Galligan Ken R. Duffy Muriel Médardx David Starobinskiy Rabia Tugce Yazicigily

2025-04-27 0 0 423.79KB 7 页 10玖币

侵权投诉

A General Security Approach for Soft-information

Decoding against Smart Bursty Jammers

Furkan Ercan†, Kevin Galligan*, Ken R. Duffy*, Muriel Médard§, David Starobinski†, Rabia Tugce Yazicigil†

†Department of Electrical and Computer Engineering, Boston University, Boston, MA, USA

§Department of Electrical Engineering and Computer Science, MIT, Cambridge, MA, USA

*Hamilton Institute, Maynooth University, Ireland

Abstract—Malicious attacks such as jamming can cause signif-

icant disruption or complete denial of service (DoS) to wireless

communication protocols. Moreover, jamming devices are getting

smarter, making them difﬁcult to detect. Forward error correc-

tion, which adds redundancy to data, is commonly deployed to

protect communications against the deleterious effects of channel

noise. Soft-information error correction decoders obtain reliabil-

ity information from the receiver to inform their decoding, but

in the presence of a jammer such information is misleading and

results in degraded error correction performance. As decoders

assume noise occurs independently to each bit, a bursty jammer

will lead to greater degradation in performance than a non-bursty

one. Here we establish, however, that such temporal dependencies

can aid inferences on which bits have been subjected to jamming,

thus enabling counter-measures. In particular, we introduce a

pre-decoding processing step that updates log-likelihood ratio

(LLR) reliability information to reﬂect inferences in the presence

of a jammer, enabling improved decoding performance for

any soft detection decoder. The proposed method requires no

alteration to the decoding algorithm. Simulation results show that

the method correctly infers a signiﬁcant proportion of jamming

in any received frame. Results with one particular decoding

algorithm, the recently introduced ORBGRAND, show that the

proposed method reduces the block-error rate (BLER) by an

order of magnitude for a selection of codes, and prevents complete

DoS at the receiver.

I. INTRODUCTION

Jammers typically aim to cause a denial of service (DoS) or

reduction of quality (RoQ) at the receiver [1] without getting

detected. They exploit the wireless transmission by mixing

their signals with legitimate communication. As a result, the

received frame becomes undecodable, which causes anoma-

lies such as increased repeat requests, reduced throughput,

prolonged delays, or a complete breakdown [2]. Powerful

jammers that blast channels with unrestrained amounts of

energy can be detected easily by the receiver. More subtle

jammers, on the other hand, might seek to inject short bursts

or lower levels of energy to disrupt communication while

circumventing their detection, causing a DoS. In general, jam-

mers must demonstrate high energy efﬁciency, low detection

probability, high levels of DoS, and resistance against physical

layer (PHY) anti-jamming techniques.

From an information-theoretic perspective, uniform jam-

mers are the most effective for reducing the channel capacity

and the code rate [3]. However, emerging techniques such as

rate-adaptation algorithms propose efﬁcient countermeasures

for such jammer attacks [4]. On the other hand, bursty

jammers [5] can be an effective approach for increasing the

block-error rate (BLER), where an adversary jams a burst of

bits in a transmitted frame. Bursty jammers become more

effective in increasing the BLER when their burst patterns

are unpredictable to the receiver. With increased BLER, the

receiver must compensate by reducing the code rate, which

sacriﬁces information throughput. Therefore it is essential to

study countermeasures to such jammer attacks.

Most traditional security approaches for wireless technolo-

gies are applied to upper layers in the protocol stack [6].

However, with the rapid growth in use cases and network

density, maintaining security for 5G-and-beyond technologies

has become a challenge [7]. PHY-layer security is an emerging

solution to threats that arise with evolving adversaries [8].

Under such adversarial behavior, machine learning-based ap-

proaches [9], [10] and spectrum sensing-based approaches [11]

have been proposed to counter jamming. Our paper speciﬁcally

focuses on jamming attacks on soft-information decoders, a

topic that has received scant attention in the literature. Our

anti-jamming approach applies to general coding schemes

and can be effortlessly supported on the physical layer with

minimal computational overhead.

In this work, we consider a smart, reactive jammer that

is bursty and only active during a fraction of the trans-

mission. It is assumed that transmission parameters, such

as the modulation and the subcarrier frequency, are known

to the adversary. To counter such an attack, we propose a

modiﬁed log-likelihood ratio (LLR) computation that takes

the conditional probability of jamming into account for each

index of the received frame. The computation of this posterior

probability is performed in two steps. First, an initial value is

calculated based on the received signal strength. Anchor points

in the received frame, for which the conditional jamming

probability is high, are then used to inform the jamming esti-

mates of neighbouring points, based on Markov state transition

probabilities. The proposed method is general to any receiver

and carried out before decoding. Simulation results show that

the proposed method unveils a signiﬁcant amount of the attack,

and therefore the attacker cannot maintain their deniability.

Using the universal ORBGRAND algorithm [12], [13], it is

shown that an order of magnitude of BLER performance can

be recovered with the proposed method and a complete DoS is

prevented, using different codebooks, i.e. random linear codes

(RLCs) and 5G cyclic redundancy check-aided Polar codes

arXiv:2210.04061v1 [cs.IT] 8 Oct 2022

(CA-Polar).

The rest of the paper is organized as follows. In Section II,

preliminaries are detailed. In Section III, the smart bursty jam-

mer model and proposed LLR approach with the conditional

jamming probability computation is presented. Section IV

explains how to approximate the conditional probability of

jamming. Results are presented in Section V, followed by

concluding remarks in Section VI.

II. PRELIMINARIES

A. PHY Jammer Models

Protection against an adversary is not possible if the ad-

versary has unlimited resources. Hence, we assume that the

adversary must operate under a set of constraints. A fully mod-

eled adversary must have assumptions, goals, and capabilities

[14]. Although there are numerous categorizations of jammers

in the literature, the PHY jammer models can be summarized

in the following two categories [2], [15].

1) Constant jammers: As their name suggests, constant

jammers continuously emit disruptive signals over the com-

munication medium. Constant jammers are primitive and often

can be detected through the radio signal strength indicator

(RSSI) component of the receiver. Simple measures such as

frequency hopping can be taken as a precaution against these

types of jammers [16]. Moreover, constant jammers are power

inefﬁcient, which limits their ability to be mobile.

2) Reactive jammers: As a power-efﬁcient and more in-

telligent alternative, reactive jammers emit signals only when

it senses a legitimate transmission taking place. This type of

jammer causes a signal collision at the receiver that disrupts

either part of or all of the frame. Prevention techniques for

these types of jammers include interference and RSS sampling

[17]. Carefully engineered, smart, reactive jammers are the

most challenging type of jammer [18].

Usually, the error correction algorithms embedded in the

PHY can be considered as a ﬁrst response against such

undesired attacks. However, as the error-correcting codes

(ECCs) are standardized, their error correction capability is

known to the adversary. Therefore, a jammer can corrupt just

enough amount of transmission to cause the decoding to fail,

eventually causing a DoS.

B. Channel model

Every soft-information decoder requires LLR as an input

which determines the hard output value of each received sig-

nal, and also acts as a measure of reliability for those signals.

In regular conditions, a larger LLR magnitude indicates more

conﬁdence in the received signal.

Let bn, a binary channel input of length n, be modulated

using binary phase-shift keying (BPSK) with the mapping

bn∈ {0,1}n→xn∈ {+1,−1}n,

where xnis the modulated channel input variable sequence.

Assuming equiprobable symbols and IID noise, given a real-

AWGN

State

AWGN &

Jamming

State

Fig. 1. Two-state Markov chain model for the reactive jammer model, with

transition probabilities band g. The state of the chain for bit iis denoted Si.

ization of the received signal, yn= (y0, y1,··· , yn−1), the

LLRs can be calculated per-bit as

L(yi|A) = 2yi

σ2

, for each i∈ {1, . . . , n},(1)

where iindicates the bit index of the received frame, the

conditioning on Aindicates it is an AWGN channel without

jamming, and σAis the standard deviation of the channel

noise.

III. EVALUATING LLRSUNDER JAMMING

A. Threat Model

The adversary is modeled as a jammer which disguises itself

by injecting zero-mean Gaussian noise into the system. It is

assumed that the smart jammer can retrieve the modulation and

subcarrier frequency of operation and therefore injects jammer

signals at the legitimate transmission frequency. In order not

to alert RSSI of the transmission system, the jammer interferes

only a fraction of the time and does so randomly in a bursty

fashion. The occurrence of jamming is modeled as a Markov

chain at the level of transmitted bits.

Fig. 1 depicts the two-state Markov chain for the jammed

channel model. The state Ais AWGN only with zero mean

and variance σ2

A. The Jstate denotes that jamming is present

in the channel, with total variance σ2

σ2

J=σ2

V+σ2

A. (2)

Here, σ2

Vis the variance of the signal introduced by the

jammer, which is an independent Gaussian random variable.

The state transitions are modeled to occur per-bit. The state

transition parameters band gdenote the probabilities of

passing from the AWGN state to the jamming state and vice

versa, respectively. The parameters b,g,σ2

J, and σ2

Acan be

estimated, and so are assumed known to the receiver.

B. LLR Calculation Under Jamming

Given that a received signal yiis certainly affected by

jamming, then its noise is independent from that impacting

other bits and the LLR would be

L(yi|J) = 2yi

σ2

(3)

instead of (1), where σ2

Jis obtained using (2). In practice,

however, the receiver does not have certainty on whether

a signal has been impacted by jamming and that induces

hidden Markov dependencies in the calculation of the LLRs.

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：

AGeneralSecurityApproachforSoft-informationDecodingagainstSmartBurstyJammersFurkanErcany,KevinGalligan*,KenR.Duffy*,MurielMédardx,DavidStarobinskiy,RabiaTugceYazicigilyyDepartmentofElectricalandComputerEngineering,BostonUniversity,Boston,MA,USAxDepartmentofElectricalEngineeringandComputerScience,MIT...

展开>> 收起<<

A General Security Approach for Soft-information Decoding against Smart Bursty Jammers Furkan Ercany Kevin Galligan Ken R. Duffy Muriel Médardx David Starobinskiy Rabia Tugce Yazicigily.pdf

共7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！

A General Security Approach for Soft-information Decoding against Smart Bursty Jammers Furkan Ercany Kevin Galligan Ken R. Duffy Muriel Médardx David Starobinskiy Rabia Tugce Yazicigily

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: