# When CSI Meets Public WiFi: Inferring Your Mobile Phone Password via WiFi Signals

> An annotated edition of this paper on Fermat's Library, where scientists and readers explain its ideas, equations and history in the margins. Annotated edition: https://fermatslibrary.com/s/when-csi-meets-public-wifi-inferring-your-mobile-phone-password-via-wifi-signals

- **Authors:** Mengyuan Li et al
- **Published:** 2016
- **Topic:** Computer Science
- **Original:** https://www.dropbox.com/s/xso11lljzhuy1sb/paper.pdf?dl=1
- **Pages:** 12
- **Annotations:** 10 comments in 10 threads
- **Featured:** Fermat's Library Journal Club
- **Annotated edition:** https://fermatslibrary.com/s/when-csi-meets-public-wifi-inferring-your-mobile-phone-password-via-wifi-signals

## Annotations

### Annotation 1 (page 1)

Beside the passage: “In this study, we present WindTalker, a novel and practical keystroke inference framework that allows an attacker to infer the sensitive keystrokes on a mobile device through WiFi-based side-channel information. WindTalker is motivated from the observation that keystrokes on mobile devices will lead to diﬀerent hand …”

**Annotation** by João Batalha (2017-02-14):

> **Side-channel attacks** are attacks based on information collected from the physical implementation of a system.
>
> Examples:
> - **Power-analysis attack** - Attack that extracts information by observing the power consumption of a hardware device such as a CPU
> - **Acoustic Attack** - Attack that extracts information from the sound produced by a system (e.g. movement of mechanical pieces of a device)

### Annotation 2 (page 1)

Beside the passage: “In this study, we present WindTalker, a novel and practical keystroke inference framework that allows an attacker to infer the sensitive keystrokes on a mobile device through WiFi-based side-channel information. WindTalker is motivated from the observation that keystrokes on mobile devices will lead to diﬀerent hand …”

**Annotation** by João Batalha (2017-02-14):

> ### Multipath Signals
> When wireless signals propagate away from the transmitter, they can be affected by reflection, refraction or diffraction. Let’s say you have two signals (A and B) that leave the transmitter at the same time. Signal A propagates directly to the receiver. Signal B bounces off a wall before reaching the receiver. Signal B will have traveled a longer distance than signal A resulting in a phase shift which could in turn cause destructive or constructive interference.
> ![multipath](http://www.cisco.com/c/dam/en/us/support/docs/wireless-mobility/wireless-lan-wlan/82068-omni-vs-direct7.gif)

### Annotation 3 (page 1)

Beside the passage: “In this study, we present WindTalker, a novel and practical keystroke inference framework that allows an attacker to infer the sensitive keystrokes on a mobile device through WiFi-based side-channel information. WindTalker is motivated from the observation that keystrokes on mobile devices will lead to diﬀerent hand …”

**Annotation** by João Batalha (2017-02-14):

> Alipay is an incredibly popular payment platform. In 2015 it processed 1 trillion dollars at an average of 100 million transactions a day. It was launched in China in 2004 by Alibaba and its founder Jack Ma.
> ![alipay](http://www.silicon.co.uk/wp-content/uploads/2016/04/library_logos_alipay_large.png)

### Annotation 4 (page 2)

Beside the passage: “public WiFi architecture without compromising the victim’s device or deploying an external device very close to the victim’s device. The victim’s device is connected to a WiFi hotspot that stealthily collects the CSI from the victim’s device by enforcing the ICMP protocol. We further adopt the directional antenna to …”

**Annotation** by João Batalha (2017-02-14):

> #### Internet Control Message Protocol (ICMP)
> ICMP is an internet protocol that is used to transmit error messages and operational information between network devices. It is widely implemented, but it is usually not employed by end-user applications to transmit data. In fact, ICMP can be used to create covert channels of communication (ICMP tunnels). [Here](https://github.com/DhavalKapil/icmptunnel) is a project that allows you to transparently tunnel your IP traffic through ICMP echo and reply packets.
>
> Diagnostic tools like `ping` and `traceroute` make use of ICMP. When network devices forward an IP datagram they decrement the Time To Live (`TTL`) field in the IP header by one. If the `TTL` reaches 0 the IP datagram is discarded, and an ICMP `Time To Live exceeded in transit` message is sent back to the source of the datagram.
>
> ICMP packets have an 8 byte header and a variable size data section.

### Annotation 5 (page 2)

Beside the passage: “on the collected CSI. Speciﬁcally, we adopt low pass ﬁlter to remove the high frequency noises and we use Principal Component Analysis (PCA) to reduce the dimensionality of the feature vectors.”

**Annotation** by João Batalha (2017-02-14):

> The main idea of **Principal Component Analysis** is that you have a feature vector that describes your data (in this case Channel State Information from a number of channels - you might have multiple antenas, which communicate at a number frequencies) and you are trying to reduce the dimensionality of that feature vector to make it easier to work with.
>
> If you wish to get a better intuition about PCA check out this [visual explanation](http://setosa.io/ev/principal-component-analysis/).

### Annotation 6 (page 3)

Beside the passage: “WiFi Standards like IEEE 802.11n/ac all support Multiple- Input Multiple-Output (MIMO) and Orthogonal Frequency Division Multiplexing (OFDM), which are expected to signiﬁcantly improve the channel capacity of the wireless system. In a system with transmitter antenna number N TX , receiver antenna number N RX and OFDM …”

**Annotation** by João Batalha (2017-02-14):

> **MIMO (multiple input, multiple output)** is a technique in wireless communications that allows the use of multiple antenas at the transmitter and the receiver in order to increase the throughput of the channel and minimize errors,

### Annotation 7 (page 3)

Beside the passage: “WiFi Standards like IEEE 802.11n/ac all support Multiple- Input Multiple-Output (MIMO) and Orthogonal Frequency Division Multiplexing (OFDM), which are expected to signiﬁcantly improve the channel capacity of the wireless system. In a system with transmitter antenna number N TX , receiver antenna number N RX and OFDM …”

**Annotation** by João Batalha (2017-02-14):

> **OFDM (Ortogonal frequency-division multiplexing)** is a method in wireless communications that encodes data on multiple carrier frequencies. Ortogonal frequencies essentially means that they do not interfere with each other. OFDM has a number of advantages such as making communication more robust against fading caused by multipath propagation.

### Annotation 8 (page 3)

Beside the passage: “Many commercial devices such as Atheros 9390 [17], Atheros 9580 [22] and Intel 5300 [8] network interface cards (NICs) with special drivers provide open access to CSI value. In this study, we adopt Intel 5300 NICs, which follows IEEE 802.11n standard [1] and can work in 2.4GHz or 5GHz. By selecting N s = 30 OFDM …”

**Annotation** by João Batalha (2017-02-14):

> The Intel 5300 NIC costs about 5 USD and the directional antenas used in the experiment cost about 20 USD. This is certainly not a costly attack.

### Annotation 9 (page 4)

Beside the passage: “In particular, WindTalker builds a Sensitive IP Pool for the interested applications or services. Take the AliPay as an example. During the payment process, it will be directed to a limited number of IP addresses, which can be obtained via a series of trials. In the experimental evaluation, it is”

**Annotation** by João Batalha (2017-02-14):

> This seems like a particularly interesting insight. Alipay is certainly not the only service that suffers from this type of “information leak". `HTTPS` does not help you here. By just looking at the IP header you can figure out when a user is doing something particularly sensitive. This is rather valuable information for an attacker!

### Annotation 10 (page 9)

Beside the passage: “In a real scenario, the distance between victim’s mobile device and AP is not ﬁxed. As shown in Fig.11(a), the recovery rate of WindTalker will decrease along with the increase of the distance. However, it is observed that, even if the distance is enlarged to 1.6m, WindTalker can still achieve 1-digit recovery rate …”

**Annotation** by João Batalha (2017-02-14):

> At first it might seem that this distance requirement would make it a lot harder to actually successfully perform such attack. However, in a real world scenario it is plausible that an attacker could set the system to focus on for instance a particular table in a coffee shop.

## Full text

Extracted from the PDF. [Annotation N] marks where each annotation above sits.

#### Page 1

When CSI Meets Public WiFi: Inferring Your Mobile Phone

Password via WiFi Signals

Mengyuan Li 1 , Yan Meng 1 , Junyi Liu 1 , Haojin Zhu 1∗ , Xiaohui Liang 2 , Yao Liu 3 and Na Ruan 1

1 Shanghai Jiao Tong University 2 University of Massachusetts at Boston

3 University of South Florida

ABSTRACT

In this study, we present WindTalker, a novel and practical keystroke inference framework that allows an attacker to infer the sensitive keystrokes on a mobile device through [Annotation 1] WiFi-based side-channel information. WindTalker is motivated from the observation that keystrokes on mobile devices will lead to diﬀerent hand coverage and the ﬁnger motions, which will introduce a unique interference to the multi-path [Annotation 2] signals and can be reﬂected by the channel state information (CSI). The adversary can exploit the strong correlation between the CSI ﬂuctuation and the keystrokes to infer the user’s number input. WindTalker presents a novel approach to collect the target’s CSI data by deploying a public WiFi hotspot. Compared with the previous keystroke inference approach, WindTalker neither deploys external devices close to the target device nor compromises the target device. Instead, it utilizes the public WiFi to collect user’s CSI data, which is easy-to-deploy and diﬃcult-to-detect. In addition, it jointly analyzes the traﬃc and the CSI to launch the keystroke inference only for the sensitive period where password entering occurs. WindTalker can be launched without the requirement of visually seeing the smart phone user’s input process, backside motion, or installing any malware on the tablet. We implemented Windtalker on several mobile phones and performed a detailed case study to evaluate the practicality of the password inference towards Alipay, the [Annotation 3] largest mobile payment platform in the world. The evaluation results show that the attacker can recover the key with a high successful rate.

Keywords

Password Inference; Channel State Information; Online Payment; Wireless Security; Traﬃc Analysis

∗ Corresponding author, Email: zhu-hj@cs.sjtu.edu.cn

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. CCS’16, October 24-28, 2016, Vienna, Austria c  2016 ACM. ISBN 978-1-4503-4139-4/16/10.. . $15.00 DOI: http://dx.doi.org/10.1145/2976749.2978397

1. INTRODUCTION

Smartphones and tablets are commonly used for performing privacy sensitive transactions of banking, payment, and social applications. Unlike stationary devices connecting to a secure network and sitting in a physically-secure space, these mobile devices are often carried by a mobile user and connected to a dynamic network environment where attackers can physically approach the target user’s device and launch various direct and indirect eavesdropping attacks. While direct eavesdropping attacks aim at directly observing the input of the target device from screen and keyboard, indirect eavesdropping attacks, a.k.a. side-channel attacks make use of side channels to infer the inputs on the target devices. Prior works [2, 3, 12, 13, 15, 16, 18, 23, 25] have shown that both types of attacks can be eﬀective in certain situations. Particularly for the side-channel attacks, it is shown that the PIN and the words entered at keyboard can be inferred from the acoustic signal at microphone [3, 12, 25], electromagnetic signal at radio antenna [2], visible light at camera [18, 23], and motion status at motion sensors [13, 15, 16]. To access the side channels, these works often assume either external signal collector devices are close to the target device (for example, 30 cm) or the sensors of the target devices are compromised to provide side channel information. However, in a mobile scenario, either assumption is hardly true and the impact of attacks is thus limited. In addition, the prior works [2, 3, 12, 13, 15, 16, 18, 23, 25] have studied the keystroke inference aiming at achieving a high inference accuracy on a series of keystrokes during a relatively-long period of time. However, the keystrokes on a mobile device are not always highly sensitive. Obviously, the eavesdropping attacker has a greater interest in obtaining the payment PIN number in a short moment than a regular typing. Therefore, the application context information also needs to be considered in the keystroke inference framework. We will show how to use application context to increase the inference eﬀectiveness.

We present WindTalker, a novel and practical keystroke inference framework that allows an attacker to infer the sensitive keystrokes on a mobile device through WiFi signals. WindTalker is motivated from the observation that the typing activity on mobile devices involves hand and the ﬁnger motions, which produce a recognizable interference to the multi-path WiFi signals from the target device to the WiFi router that connects to the device. Unlike prior sidechannel attacks or traditional CSI based gesture recognition, WindTalker neither deploys external devices close to the tar-

1068

#### Page 2

get device nor compromises any part of the target device; instead, WindTalker setups a ‘rogue’ hotspot to lure the target user with free WiFi service, which is easy-to-deploy and diﬃcult-to-detect. As long as the target device is connected to the hotspot, WindTalker at the hotspot intercepts the traﬃc and time-adaptively collect the channel state information (CSI) between the target device and the hotspot.

The design of WindTalker faces three major technical challenges. i) The impact of the hand and ﬁnger movement of keystrokes on CSI waveforms is very subtle. An eﬀective signal analysis method is needed to analyze keystrokes from the limited CSI. ii) The prior CSI collection method requires two WiFi devices, one as a signal sender and the other as a signal receiver, which are deployed close to the victim. A more ﬂexible and practical CSI collection method is highly desirable for the mobile device scenario. iii) The key inference must be done at some selective moments for obtaining a sensitive keystroke, such as payment PIN number. Such context-oriented CSI collection has not been addressed by prior works. In this paper, We introduce a novel CSI based keystroke inference framework, which consists of four specifical contributions.

• We present a practical CSI collection method using

public WiFi architecture without compromising the victim’s device or deploying an external device very close to the victim’s device. The victim’s device is connected to a WiFi hotspot that stealthily collects the [Annotation 4] CSI from the victim’s device by enforcing the ICMP protocol. We further adopt the directional antenna to eliminate CSI noises introduced by other factors in public places, such as other people’s movement.

• We propose a keystroke recognition algorithm based

on the collected CSI. Speciﬁcally, we adopt low pass ﬁlter to remove the high frequency noises and we use Principal Component Analysis (PCA) to reduce the [Annotation 5] dimensionality of the feature vectors.

• We propose a context-oriented CSI collection method,

which employs both of the traﬃc analysis towards meta data in WiFi traﬃc and CSI data analysis to recognize the PIN input moment based on certain CSI tags. The proposed method can be used to successfully ﬁgure out the time of the PIN entry on Alipay (a popular mobile payment platform in China) and launch the keystroke recognition accordingly.

• We perform an extensive evaluation on keystroke in-

ference towards PIN input at the mobile payment process, which is secured by the HTTPS protocol and thus traditionally believed to be secure. Through our evaluation, we demonstrate that the attacker can infer the PIN number at a high successful rate.

To the best of our knowledge, this is the ﬁrst work to launch the keystroke inference towards PIN entry at the mobile payment (e.g., Alipay). The remainder of this paper is organized as follows. In Section 2, we introduce the background of this work. In Section 3, we introduce the research motivation by showing the correlation of keystroke and CSI changing. We present the detailed design in Section 4, which is followed by Evaluation, Real-world experiment, Discussion and Related work in Section 5, 6, 7 and 8, respectively. Finally, we give the conclusion and future work in Section 9.

RX T X

Smart device Attacker (a) IKI Model

T X

RX WiFi Router

Keyboard Attacker (b) OKI Model

Figure 1: WiFi-based Keystroke Inference Models

2. BACKGROUND

In this section, we introduce the scenario, the overview of the keystroke inference methods, and preliminaries of channel state information.

2.1 Scenario

We consider a scenario where a user has a mobile device, such as a smartphone, or a tablet and he or she is using the public free WiFi through the device. It is a very common situation that people could have in the shopping mall, the airport, and restaurants. A WiFi hotspot is set up at a corner or on the ceiling, an unnoticeable location from the user’s view. The user searches all the available WiFi signals at her device, and may choose to use the WiFi network if the name of the network “looks” good and the network is authentication-free. With the application layer security (HTTPs), the user may believe that the Internet traﬃc is protected from end-to-end such that the content shown at the device and the user’s inputs at the device will be only available to herself and the service provider. However, as we will show, our WindTalker framework suggests eﬀective keystroke inference methods targeting at the mobile device.

2.2 In-band keystroke inference model

WindTalker chooses In-band keystroke inference (IKI) model. As shown in Fig.1(a), WindTalker deploys one Commercial Oﬀ-The-Shelf (COTS) WiFi device close to the target device, which could be a WiFi hotspot. The WiFi hotspot provides free WiFi networks for nearby users. When a user connects her device to the hotspot, the WiFi hotspot is able to monitor the application context by checking the pattern of the transmitted packets. In addition, the WiFi hotspot periodically sends ICMP packets to obtain the CSI information from the target device. With the meta data of the WiFi traﬃc, the hotspot knows when the sensitive operations happen. And then, the hotspot adaptively launches CSI-based keystroke inference method to recognize sensitive key inputs. To the best of our knowledge, the IKI method we propose is the ﬁrst one using existing network protocols of IEEE 802.11n/ac standard to obtain the application context and the CSI information at mobile devices.

Note that the existing works about CSI based gesture recognition choose another strategy: Out-of-band keystroke inference (OKI) model[2]. In this model, the adversary deploys two COTS WiFi devices close to the target device and makes sure the target device is placed right between two COTS WiFi devices. One is the sender device continuously emitting signals and the other one is the receiver device continuously receiving the signals. The keystrokes are inferred from the multi-path distortions in signals.

Compared with OKI model, the proposed IKI model has the below advantages. Firstly, compared with OKI model,

1069

#### Page 3

IKI model does not require the placement of both sender andreceicerdeviceandcanbedeployedinamoreﬂexible and stealthy way. Secondly, OKI model fails to diﬀerentiate the non-sensitive operations on mobile devices (e.g., clicking the screen to open an APP or just for web-browsing) from sensitive operation (e.g., inputting the password). Instead, IKImodelallowstheattackertoobtainbothofun-encrypted meta data traﬃc as well as the CSI data to launch a more ﬁne-grained attack.

2.3 Channel State Information

The basic goal of WindTalker is measuring the impact of hand and ﬁngers’s movement on WiFi signals and leveraging correlation of CSI and the unique hand motion to recognize PIN. In the below, we brieﬂy introduce the CSI related backgrounds.

WiFi Standards like IEEE 802.11n/ac all support Multiple- [Annotation 6] Input Multiple-Output (MIMO) and Orthogonal Frequency Division Multiplexing (OFDM), which are expected to sig- [Annotation 7] niﬁcantly improve the channel capacity of the wireless system. In a system with transmitter antenna number N TX , receiver antenna number N RX and OFDM subcarriers number N s ,systemwilluseN TX × N RX × N s subcarriers to transmit signal at the same time.

CSI measures Channel Frequency Response (CFR) in different subcarriers f.CFRH (f, t) represents the state of wireless channel in a signal transmission process. Let X (f,t) and Y (f,t) represent the transmitted and received signal with diﬀerent subcarrier frequency. H (f,t)canbecalculatedinreceiverusingaknowntransmittedsignalvia

H (f, t)= Y (f,t) X (f, t)

Since the received signal reﬂects the constructive and destructive interference of several multi-path signals scattered from the wall and surrounding objects, the movements of the ﬁngers while password input can generate a unique pattern in the time-series of CSI values, which can be used for keystrokes recognition.

Many commercial devices such as Atheros 9390 [17], Atheros 9580 [22] and Intel 5300 [8] network interface cards (NICs) [Annotation 8] with special drivers provide open access to CSI value. In this study, we adopt Intel 5300 NICs, which follows IEEE 802.11n standard [1] and can work in 2.4GHz or 5GHz. By selecting N s = 30 OFDM subcarriers, Intel 5300 NICs collect CSI value for each TX-RX antenna pair.

3. MOTIVATION

In this section, we illustrate the rationale behind CSI based keystroke inference on smart phones using real-world experiments. Fig.2(a) shows the sketch of typical touching screen during the PIN entry for mobile payment (e.g., Alipay or Wechat pay). We particularly focus on the vertical touch and the oblique touch, which are two most common touching gestures [4, 7, 20]. As shown in the left of Fig.2(b), oblique touch is the most common typing gesture, which happens when people press diﬀerent keys. Vertical touch usually happens when the human continuously presses the same key, (e.g., continuously pressing 1) in the right of Fig.2(b).

We further investigate how these two common typing gestures inﬂuence CSI. Generally speaking, since CSI reﬂects the constructive and destructive interference of several multipath signals, the change of multi-path propagation during

(a) Finger typing

X

Y

Z

X

Y Z

(b) Click

X

Y

Z

X

Y

Z

(c) Coverage

Figure 2: Finger’s inﬂuence on CSI

2000 4000 6000

15

20

25

30

35

40

Sam p le

CSI Amplitude

Number 1

4 4.2 4.4

x 10 4

0

5

10

15

20

Sam p le

CSI Amplitude

Number 6

5.6 5.8 6 6.2

x 10 4

15

20

25

30

35

Sam p le

CSI Amplitude

Number 8

7.3 7.4 7.5 7.6 7.7

x 10 4

15

20

25

30

35

40

Sample

CSI Amplitude

Number 0

(a) Continuously Click in Diﬀerent Keys

0 1 2 3 4 5 6 7 8 x 10 4

0

10

20

30

40

50

Sample

CSI Amplitude

(b) Continuously Click in the Same Key

Figure 3: CSI Change When Typing

the PIN entry can generate a unique pattern in the timeseries of CSI values, which can be used for keystrokes inference. From our experiments, we found that two main factors contributing to CSI changes are hand coverage and the ﬁnger click.

Hand coverage and ﬁnger position on a smart phone touchscreen are one of the major factors that cause the ﬂuctuation of CSI waveform. It is widely acceptable that ﬁnger position and coverage have a direct impact on the calling quality. Similarly, since time series of CSI waveform reﬂects the interference of several multi-path signals, diﬀerent ﬁnger position and coverage will inevitably introduce the interference to the WiFi signals and thus lead to the changes of the CSI. We further demonstrate the it via a series of experiments. Fig.3(b) shows a CSI stream when continuously pressing diﬀerent number from 1 to 9, followed by 0, each for 5 times. It can be seen that the diﬀerent coverages lead to the diﬀerent ﬂuctuation range of the CSI value, which can be exploited for key inference.

Finger click is another important factor that contributes to the ﬂuctuation of CSI. Compared with CSI change caused

1070

#### Page 4

by the hand coverage, the experiment shows that ﬁnger click has a more direct inﬂuence on CSI by introducing a sharp convex in Fig.3(a), which corresponds to the quick click’s inﬂuence on multi-path propagation. This feature can be used to distinguish the oblique touches in the case that the human continuously presses the same key or the adjacent keys, which produce similar CSI values.

4. THE DESIGN OF WINDTALKER

4.1 System Overview

The basic strategy of WindTalker is hitting two birds with one stone. On one hand, it analyzes the WiFi traﬃc to identify the sensitive attack windows (e.g., PIN number) on smartphones. On the other hand, as long as an attack window is identiﬁed, WindTalker starts to launch the CSI based keystroke recognition. As shown in Fig.4, WindTalker is consisted of the following modules: Sensitive Input Window Recognition Module, which is responsible for distinguishing the sensitive input time windows, ICMP Based CSI Acquirement Module, which collects the user’s CSI data during his access to WiFi hotspot, Data Preprocessing Module,which preprocesses the CSI data to remove the noises and reduce the dimension, Keystroke Extraction Module, which enables WindTalker to automatically determine the start and the end point of keystroke waveform, and Keystroke Inference Module, which compares the diﬀerent keystroke waveforms and determines the corresponding keystroke.

Connected

Victim

Sensitive Input

Directional Antenna

Internet Control

Message Protocol

CSI

Noise Removal

Dimension Reduction

Keystroke Extraction

Keystroke Recognition

Output

WiFi Packets

Analysis

Figure 4: WindTalker Framework

4.2 Sensitive Input Window Recognition Mod-

ule To distinguish the time window of the sensitive input from that of insensitive input, WindTalker captures all the packets of the victim with Wireshark and records timestamp of each CSI data. Currently, most of the important applications are secured via HTTPS, which provides end-to-end encryption and prevents the eavesdropper from obtaining the sensitive data such as the password. Our insight is that though HTTPS provides end-to-end encryption, it cannot protect the meta data of the traﬃc such as the IP address of the destination sever, which can be used to recognize sensitive input window.

In particular, WindTalker builds a Sensitive IP Pool for the interested applications or services. Take the AliPay as an example. During the payment process, it will be directed to a limited number of IP addresses, which can be obtained [Annotation 9] via a series of trials. In the experimental evaluation, it is

shown that, for Alipay users, the traﬃcs of the users under thesamenetworkwillbedirectedtothesameserverIP, which will last for a period (e.g., several days for one round of experiment). This allows WindTalker to ﬁgure out the sensitive input time window.

During the attack process, as long as the traﬃc to the Sensitive IP Pool is observed, WindTalker will record the corresponding start time and the end time, which serve as the start and the end of the Sensitive Input Window. Then, it starts to analyze the CSI data in this period to launch the password inference attack via WiFi signals.

4.3 ICMP based CSI Acquirement Module

4.3.1 Collecting CSI Data by Enforcing ICMP Reply

Diﬀerent from the previous works which rely on two devices including both of the sender and the receiver to collect CSI data, we apply an approach that leverages Internet Control Message Protocol (ICMP) in hotspot to collect CSI data during the user accesses to the pre-installed access point. In particular, WindTalker periodically sends a ICMP Echo Request to the victim smartphone, which will reply an Echo Reply for each request. To acquire enough CSI information of the victim, WindTalker needs to send ICMP Echo Request at a high frequency, which enforces the victim to replay at the same frequency. In practice, WindTalker can work well for several smartphones such as XiaoMi, Samsung and Nexus at the rate of 800 packets per second. It is important to point out that this approach does not require any permission of the target smartphone and is diﬃcult to be detected by the victim.

ICMP based CSI collection approach introduces a limited number of extra traﬃc. For a 98 bytes ICMP packet, when we are sending 800 ICMP packets per second to the victim, it needs only 78.4 kB/s for the attack where 802.11n can theoretically support the transmission speed up to 140 Mbits per second. It is clear that the proposed attack makes little interference to the WiFi experience of the victim.

4.3.2 Reducing Noise via Directional Antenna

CSI will be inﬂuenced by both ﬁnger movement and people’s body movement. One of the major challenges of obtaining the exact CSI data in public space is how to minimize the interference caused by the nearby human beings. We present a noise reduction approach by adopting the directional antenna. Diﬀerent from omni-directional antennas that have a uniform gain in each direction, directional antennas have a diﬀerent antenna gain in each direction. As a result the signal level at a receiver can be increased or decreased simply by rotating the orientation of the directional antenna. WindTalker employs directional antenna to focus the energy toward the target of interest, which is expected to minimize the eﬀects of the nearby human body movement.

WindTalker employs a TDJ-2400BKC antenna working in 2.4GHz to collect CSI data of the targeted victim, whose Horizontal Plane -3dB Power Beamwidth and Vertical Plane -3dB Power Beamwidth are 30 ◦ and 25 ◦ respectively. Considering the case that the distance between the victim and access point is 1.5 meter, we illustrate the eﬀective acceptance area of 0.67 meter high and 0.80 meter long.

Fig.5 shows the comparison of CSI collection with directional antenna and without directional antenna in public place. Fig.5(b), Fig.5(c), Fig.5(d) show CSI amplitude in

1071

#### Page 5

0 2000 4000 6000 20

25

30

35

40

45

50

Sample

CSI Amplitude

(a) Omni-directional Antenna in 75cm

0 2000 4000 16

18

20

22

24

26

28

Sample

CSI Amplitude

(b) Directional Antenna in 75cm

0 2000 4000 6000 18

20

22

24

26

28

Sample

CSI Amplitude

(c) Directional Antenna in 125cm

0 2000 4000 6000 16

18

20

22

24

Sample

CSI Amplitude

(d) Directional Antenna in 150cm

Figure 5: Antenna Performance in Public Place

the case that a victim is located at 75, 125, 150 cm accordingly away from directional antenna while one people moving nearby. Unique pattern caused by ﬁnger click in number 1 can be easily caught from the original CSI stream without any preprocessing. However, these patterns are submerged in human body’s inﬂuence on CSI stream obtained by omnidirectional antenna even when the victim and attacker is close as 75 cm, which is shown in Fig.5(a).

4.4 Data Preprocessing Module

Before launching keystroke inference module, WindTalker needs to preprocess the CSI data to remove the noises introduced by commodity WiFi NICs due to the frequent changes in internal CSI reference levels, transmit power levels, and transmission rates. To achieve this, WindTalker ﬁrst turns to low pass ﬁlter to remove the high frequency noise. Then, WindTalker leverages the Principal Component Analysis to reduce the dimensionality of the feature vectors to enable better analysis of the data.

4.4.1 Low Pass Filtering

The observation behind low pass ﬁltering is that the variations of CSI waveforms caused by ﬁnger motion lie at the low end of the spectrum while the frequency of the noise lies at the high end of the spectrum. To remove noise, we adopt Butterworth low-pass ﬁlter, which is designed to have a ﬂat frequency response in the passband and thus does not distort the ﬁnger motion signal much. It is observed that the frequencies of the variations in CSI time series due to hand and ﬁnger movements lie between 2 Hz and 30 Hz. As we sample CSI values at a rate of S = 800 packets/s, WindTalker sets some parameters to choose a proper ﬁlter in which the transition band ranges from 30Hz to 80Hz. We set the passband corner frequency W p = 2∗f p S = 2∗30 800 ≈ 0.075 πrads/sample with 1 corresponding to the normalized Nyquist frequency and stopband corner frequency W s = 2∗f s S = 2∗80 800 ≈ 0.2 πrads/sample. Passband ripple in decibels is 1 and Stop-

band attenuation in decibels is 40. After low-pass ﬁlter, most of the burst noises can be removed.

4.4.2 Dimension Reduction

Dimension reduction is essential for keystroke inference via CSI information. For a CSI recording system using Intel 5300 NICs with N TX transmitter antennas and N RX receiver antennas, it can collect N TX ×N RX ×30 CSI streams. It is important to reduce the dimensionality of the CSI information obtained from 30 subcarriers in each TX-RX stream and recognize those subcarriers which show the strongest correlation with the hand and ﬁnger movements. WindTalker adopts PCA, which is expected to choose the most representative or principal components from all CSI time series. PCA is also expected to remove the uncorrelated noisy components. The procedure of dimension reduction of CSI time series based on PCA includes the following steps.

Sample Centralization: Performing sample centralization in every subcarries. We use a matrix H to present originalCSIstreamdata. Forexample,inasystemwithone pair of TX-RX antenna, we will get 30 CSI streams from 30 subcarriers. Thus, with sample rate S and time T , H has dimension of M × 30, where M = S × T .Everycolumn of H represents a CSI time series data stream in one subcarrier. Then we calculate the mean value of each column in H and subtract the corresponding mean values in every column. After the centralization step, we get a processed matrix H p . Calculating Covariance Matrix: Calculating the correlation matrix of H p as H p T × H p . Calculating Eigenvalues and Eigenvectors of Covariance: Calculating the Eigenvalues and Eigenvectors of Covariance. The Eigenvectors are normalized to unit vectors.

Choosing Main Eigenvalues: Sorting the Eigenvalues from large to small and choosing the maximum k number of Eigenvalues. Then the corresponding k Eigenvectors are used as the column vectors to form a Eigenvector matrix. We will get a Eigenvector matrix whose dimension is 30×k.

Data Reconstruction: Projecting H p onto the selected k Eigenvector matrix. The reconstruction CSI data stream H r has the dimension of M × k.

H r (M × k)=H p (M × 30) × EigenV ectors(30 × k)

With PCA, we can identify the most representative components inﬂuenced by the victim’s hand and ﬁnger’s movement and remove the noisy components at the same time. In our experiment, it is observed the ﬁrst k =4components almost show the most signiﬁcant changes in CSI streams and the rest components are noises. We only take one PCA component from the ﬁrst 4 components in the password inference module. We observed that the ﬁrst PCA component reserves most changes in CSI while the ambient noise is weakly. Otherwise, the ﬁrst component has a large noise and the succeeding k − 1 components reserve most changes in CSI.

4.5 Keystroke Inference Module

4.5.1 Keystroke Extraction

By processing the low pass ﬁltering and dimension reduction, it is observed that the CSI data shows a strong correlation with the keystrokes. In the experiment, the sharp rise

1072

#### Page 6

0 2000 400 0 6000 8000 10000 12000 14000 16000

-1

0

1

2

3

Sam p le

CSI Value

(a) After Once Filter

0 2000 4000 6000 8000 10000 12000 14000 16000

-1

0

1

2

3

Sample

CSI Value

(b) After Twice Filter

1.3 1.4 1.5 1.6 1.7

x 10 4

−0.5

0

0.5

1

1.5

2

Sample

CSI Value

J Anchor Points

(c) Judgement Value

0 2000 4000 6000 8000 10000 12000 14000 16000

-1

0

1

2

3

Sample

CSI Value

CSI Value Variance

(d) Variance Scan

0 2000 4000 6000 8000 10000 12000 14000 16000

-1

0

1

2

3

Sample

CSI Value

CSI Value Start Point End Point

(e) The Results of Extraction

1.4 1.5 1.6 1.7

x 10 4

−0.5

0

0.5

1

1.5

2

Sample

CSI Value

J Start Point End Point

(f) Keystroke Area

Figure 6: Keystroke Extraction

andfalloftheCSIwaveformsignalsareobservedincoincidence with the start and end of ﬁnger touch. How to determine the start and the end point of CSI time series during a keystroke is essential for keystroke recognition. However, the existing burst detection schemes such as Mann-Kendall test [9], moving average method [10] and cumulative anomalies [14] do not work well in our situation since the CSI waveform has many change-points during the password input period. Therefore, we propose a novel detection algorithm to automatically detect the start and end point. The proposed algorithm includes the following three steps.

Waveform Proﬁle Building: AsshowninFig.6(a),itis observed that there is a sharp rise and fall which correspond to the ﬁnger motions. However, there is a strong noise which prevents us from extracting interested CSI waveform related to the keystrokes. This motives us to perform another round of noise ﬁltering. In the experiment, we adopt Butterworth ﬁlter and choose 10Hz as the cutoﬀ frequency to make the waveform smooth. After being ﬁltered, the CSI data during the keystroke period are highlighted while the waveform during non-keystroke period becomes smooth, which are shown in Fig.6(b).

CSI Time Series Segmentation and Feature Segment Selection: To extract the CSI waveforms for individual keystrokes, we slice the CSI time series into multiple segments, which be grouped together according to the temporal proximity, and then choose the center of segment as the feature waveform for a speciﬁc keystroke. Without loss of the generality, it is assumed that each segment contains W samples. Given the sampling frequency S, and the time duration τ , W can be represented by S × τ.Forthewaveform with time duration of T , the number of segments N can be calculated as below:

N =



T × S

W



It is observed that the CSI segments during the keystroke period show a much larger variance than those happening out of the period, which is shown in Fig.6(d). Motivated by this, we are only interested in the segments with the variance which is larger than a predetermined threshold while

removing the segments with the variance under this threshold. The selected segments are grouped into various groups according to the temporal proximity (e.g., ﬁve adjacent segments grouped into one group in the practice). Each group represents the CSI waveform of an individual keystroke and the center point of this group is selected as the feature segment of this keystroke. The process of time series segmentation and feature segment selection is shown in Fig.6(d).

Keystroke Waveforms Extraction: To extract keystroke waveforms, the key issues is how to determine the start and the end point of CSI time series, which could cover as much keystroke waveform as possible while minimizing the coverage of the non-keystroke portion. We choose the average value of the segment samples J as the key metric and the intersection of J and the CSI waveform serves as the anchor points. In particular, starting from the leftmost anchor point, it performs a local search and chooses the nearest local extremum which is below J as the start point. Similarly, beyond the rightmost anchor point, it can choose the nearest local extremum which is below J as the end point. As shown in Fig.6(c), Fig.6(f), Fig.6(e), with the start and the end point, keystroke waveform can be extracted.

Thus, we can divide a CSI stream into several keystroke waveform. The i th keystroke waveform K i from the k th

principal component H r (:,k) of CSI waveforms as follows.

K i = H r (s i : e i ,k)

where s i and e i be the start and the end time of i th keystroke. After keystroke extraction, we use these keystroke waveform to conduct recognition process.

4.5.2 Keystroke recognition

One of the major challenges for diﬀerentiating keystrokes is how to choose the appropriate features that can uniquely represent the keystrokes. As shown in Fig.7, it is observed that diﬀerent keystrokes will lead to diﬀerent waveforms, which motivates us to choose waveform shape as the feature for keystroke classiﬁcation. To compare the waveforms of diﬀerent keystrokes, we adopt the Dynamic Time Warping (DTW) to measure the similarity between the CSI time series of two keystrokes. However, directly using the keystroke

1073

#### Page 7

0 500 1000 1500 2000 −10

0

10

20

Sample Index

CSI Value

0 500 1000 1500 −5

0

5

10

15

Sample Index

CSI Value

(a) Two samples of keystroke waveforms number 2

0 500 1000 1500 −15

−10

−5

0

5

Sample Index

CSI Value

0 500 1000 1500 2000 −10

−5

0

5

10

Sample Index

CSI Value

(b) Two samples of keystroke waveforms number 4

Figure 7: CSI Diﬀerence Between Two Number

waveforms as the classiﬁcation features leads to high computational costs in the classiﬁcation process since waveforms contain many data points for each keystroke. Therefore, we leverage Discrete Wavelet Transform (DWT) to compress the length of CSI waveform by extracting the approximate sequence. In the below, we will introduce the details.

4.5.3 Discrete Wavelet Transform

Diﬀerent from the traditional frequency analysis such as Fourier Transform, DWT is the time-frequency analysis which has a good resolution at both of the time and frequency domains. A discrete signal x [n] can be expressed in terms of the wavelet function by the following equation:

x[n]= 1 √

L



k

W φ [j 0 ,k]φ j 0 ,k [n]+ 1 √

L

∞ 

j=j 0



k

W ψ [j, k]ψ j,k [n],

where x[n] represents the original discrete signal and L represents the length of x[n]. φ j 0 ,k [n]andψ j,k [n]referto wavelet basis. W φ [j 0 ,k]andW ψ [j, k] refer to the wavelet coeﬃcients. The functions φ j 0 ,k [n] refer to scaling functions and the corresponding coeﬃcients W φ [j 0 ,k] refer to the approximation coeﬃcients. Similarly, functions ψ j,k [n] refer to wavelet functions and coeﬃcients W ψ [j, k] refer to detail coeﬃcients. To obtain the wavelet coeﬃcients, the wavelet basis φ j 0 ,k [n]andψ j,k [n] are chosen to be orthogonal to each other.

During the decomposition process, the origin signal is ﬁrst divided into the approximation coeﬃcients and detail coefﬁcients. Then the approximation coeﬃcients are iteratively divided into the approximation and detail coeﬃcients of next level. The approximation and the detail coeﬃcients in j th

level can be calculated as follows:

W φ [j 0 ,k]=x[n],φ j 0 +1,k [n] = 1 √

L



n

x[n]φ j 0 +1,k [n]

W ψ [j, k]=x[n],ψ j+1,k [n] = 1 √

L



n

x[n]ψ j+1,k [n]

In the ﬁrst DWT decomposition step, the length of approximation coeﬃcients is half of L.Forthej th decomposition step, the length is half of the previous decomposition. We use the approximation coeﬃcients to compress the original keystroke waveforms to reduce computational cost. In

order to achieve the tradeoﬀ between the sequence length reducing and preserving the waveform information, we need to choose an appropriate wavelet basis and decomposition level. In practice, we choose Daubechies D4 wavelet and perform 3-level DWT decomposition in the classiﬁcation model. Therefore, for i th keystroke, the third level approximation coeﬃcients of K i is chosen as the feature of the keystroke.

4.5.4 Dynamic Time Warping

To compare features of diﬀerent keystrokes, WindTalker adopts DTW to achieve keystroke recognition. DTW utilizes dynamic programming to calculate the distance between two time series of keystroke waveforms with diﬀerent lengths. With DTW, the sequences (e.g., time series) are warped nonlinearly in the time dimension to measure their similarity. The input of DTW algorithm is two time series and the output is the distance between two series. A low distance indicates that these two sequences are highly similar.

4.5.5 Classiﬁer Training

We build a classiﬁer to recognize the keystrokes based on their keystroke waveform shapes. Our classiﬁer gives each keystroke waveform a set of scores, which allows the keystrokes to be diﬀerentiated based on the user’s training dataset (keystrokes on diﬀerent numbers). For a certain key number, classiﬁer ﬁrst calculates the DTW distances between the input waveform and all the key number’s waveforms in dataset. Then classiﬁer chooses the average value of the previous distances as the score between the input waveform and the certain key number. The smaller the score, the higher possibility the certain number is actual input number. The classiﬁer choose the key number which has the minimum score as the predicted key number. Note that the classiﬁer saves all scores in order to generate password candidates in Section 5.3.

5. EVALUATION

5.1 System Setup

WindTalker is built with the oﬀ-the-shelf hardware, which is actually a commercial laptop computer equipped with Intel 5300 NIC with one external directional antenna and two omni-directional antennas. WindTalker also serves as the WiFi hotspot to attract the users to access to the WiFi. The laptop runs Ubuntu 14.04 LTS with a modiﬁed Intel driver to collect CSI data. To collect the CSI data related to the user’s touch screen clicks, WindTalker uses ICMP echo and reply to achieve the sampling rate of 800 packets/s. In this evaluation, the distance between the mobile user and the AP is 75 cm and the AP is placed on the left side of mobile phone.

In the online phase, we recruit 10 volunteers to join our evaluation, including 7 males and 3 females. All of the volunteers are right-handed and they perform the touch-screen operations by following their own fashions. During the experiment, the volunteers should participate in the data training phase and keystroke recognition phase by inputting the numbers according to the system hints. In the data training phase, WindTalker records each input and its corresponding CSI data. In the test phase, WindTalker infers the input data based on the observed CSI time series. The training data and testing data collection should be ﬁnished within 30

1074

#### Page 8

Figure 8: Classiﬁcation Accuracy per key

minutes since CSI may change with the change of environment.

We start the evaluation by testing the classiﬁcation accuracy and the 6-digit password inference accuracy. Then we investigate various metrics that may inﬂuence the inference accuracy of WindTalker including the distance and the direction. Afterwards we perform a more speciﬁc case study by inferring the password of mobile payment for Alipay in Section 6. In the current stage evaluation, we only perform user speciﬁc training and will discuss it’s limitation in Section 7.

5.2 Classiﬁcation Accuracy

In Section 3, we have shown that diﬀerent keystrokes may be correlated with diﬀerent CSI waveforms. In this section, we aim to explore whether the diﬀerences of keystroke waveforms are large enough to be used for recognizing diﬀerent keys inputs in the real-world setting. We collected training and testing data from 10 volunteers. Each volunteer ﬁrst generates 10 loop samples, where a loop is deﬁned as the CSI waveform for key number from 0 to 9 by pressing the corresponding digit. After that, we evaluate the classiﬁcation accuracy of WindTalker through the collected CSI data. The classiﬁcation accuracy is evaluated in terms of cross validation accuracy. In our problem setting, for every 10 loops dataset, we pick up one loop in turn for the testing data and choose the other 9 loops as the training dataset. WindTalker adopts the classiﬁer introduced in Section 4.5 to recognize the keystroke. We perform the evaluation on Samsung Note 5 , Xiaomi Redmi Note 3 and Nexus 5. These mobile are run with Android 6.0.1, 5.0.2 and 6.0.1, respectively. When we use all ten loops data, WindTalker achieves average accuracy classiﬁcation of 81.8% in Xiaomi, 73.2% in Nexus and 64% in Samsung. Fig.8 shows average classiﬁcation accuracy of all 10 volunteers in 10 PIN number.

Fig.9 describes the color map of confusion matrix of Xiaomi. For a speciﬁc typed number, it gives the corresponding inference results. The darker the area is, the higher the possibility of keystroke inference result is. Intuitively, it is easier for an input number that is confused with the neighboring numbers during the keystroke inference process.

5.3 Password Inference

In a practical scenario setting, it may not be easy for WindTalker to get 10 training samples for each PIN number. So in the remaining section, we only use 3 samples per number for training. To illustrate the performance of WindTalker for password Inference, in this part, we ask volunteers to press 10 randomly generated 6-digit passwords

Predicted Key Number

Actual Key Number

1 2 3 4 5 6 7 8 9 0

1 2 3 4 5 6 7 8 9 0

Figure 9: Color Map

Table 1: Recovery Rate and Candidates Number

Phone One Two Three SamSung 0.63 0.83 0.89 XiaoM i 0.79 0.88 0.95

and use their corresponding 3 loops as training dataset. This experiment is repeated in both Samsung and XiaoMi.

We test totally 200 set of passwords, which include 1200 keys. The inference results show that totally 852 keys were recovered. As shown in Table.1, WindTalker can achieve an average 1-digit recovery rate of 79.0% in XiaoMi and 63.0% in SamSung. For a 6-digit password in AliPay, the attacker can try several times to recover the password at an increased successful rate. Thus, we introduce a new metric, recovery rate with Top N candidates, which indicates the rate of successfully recovering the password for trying N times and represents a more reasonable metric to describe the capability of the attacker in the practical setting. As shown in Table.1, if we evaluate the 1-digit recovery rate under top 2 and top 3 candidates, it is found that the recovery rate can be signiﬁcantly improved.

We further study how many candidates can help us to succeed in predicting the right 6-digit payment password in WindTalker. In particular, we will investigate the inference accuracy under top N candidates. In the experiment, each 6-digit password will be correlated with six CSI waveforms. For each waveform, WindTalker calculates the probability of matching the waveform with the predict key number. The probability of a predicted password is deﬁned by the product of the six probabilities. For a 6-digit password, we can obtain 100000 predicted password, then sort these password by their probabilities in descending order. A successful password inference is deﬁned as that the real password

0 5 10 15 20 0

10

20

30

40

50

60

Number of candid ate passwords Password Inference Accuracy (%)

(a) Top 20 Candidates

20 40 60 80 100 50

60

70

80

90

Number of candidate p asswords Password Inference Accuracy (%)

(b) Top 100 Candidates

Figure 10: 6-digit Password Inference Accuracy

1075

#### Page 9

(a) Distance

0 50 100

−20

−10

0

10

20

30

Temporal Units

Approximation Coefficients

75 cm

0 50 100

−10

0

10

20

Temporal Units

Approximation Coefficients

100 cm

0 50 100

−20

−10

0

10

20

30

Temporal Units

Approximation Coefficients

125 cm

0 100 200 −30

−20

−10

0

10

20

30

Temporal Units

Approximation Coefficients

150 cm

PCA 2nd Component PCA 3rd Component PCA 4th Component

(b) CSI Shape Change by Distance

Figure 11: Distance’s inﬂuence in WindTalker

is included in top N candidates. In Fig.10(a) ,we give the password inference accuracy under top N candidates, where N ranges from 1 to 20. The result is encouraging. It is shown that, given top 1 candidate, the inference accuracy is only 20%. The inference rate can be signiﬁcantly improved if given top 5 candidates or top 10 candidates, which correspond to 38% and 42%, respectively. It is also shown in Fig. 10(b) that, if given enough top N candidates (e.g., set N as 85), the inference accuracy can reach almost 80%.

5.4 Impact of Distance and Direction

There are many factors potentially impacting the CSI. Even clicking at the same key, the diﬀerent distance and direction between AP and the mobile device may also lead to a quite diﬀerent CSI. We will investigate the impact of the distance and the direction on CSI in our experiments.

5.4.1 Distance

In a real scenario, the distance between victim’s mobile device and AP is not ﬁxed. As shown in Fig.11(a), the recovery rate of WindTalker will decrease along with the increase of the distance. However, it is observed that, even if the distance is enlarged to 1.6m, WindTalker can still achieve 1-digit recovery rate 70% under top 3 candidates. It demonstrates that WindTalker can work well even if the distance reaches 1.6m. This is because with WindTalker, the [Annotation 10] attacker can enforce the smart phones to send WiFi signals and AP to receive the WiFi signals. In this setting, though the distance between the victim and receiver (AP) is en-

Figure 12: Accuracy in Diﬀerent Direction

larged, the distance between the WiFi sender (smart phone) and the victim (ﬁngers) is relatively stable, which guarantees key recognizing. Fig.12 shows that both CSI shape and degree will change under diﬀerent distance. This indicates that WindTalker needs to retrain dataset even for the same victim with diﬀerent distances. To partially solve this limitation, in practice, the attacker can ﬁx the location of table and chairs, which will make the user’s position relatively stable.

5.4.2 Direction

The relative direction between the victim and attacker may seriously aﬀect the CSI since diﬀerent directions mean diﬀerent multi-path propagation between the transmitter and the receiver. Thus, we show the performance of WindTalker under diﬀerent directions. Note that the mobile device is in front of victim in experiments. It is important to point out that, for a right-handed user, WindTalker shows a better performance when the AP is on the left side of the victim. This is because it is easier for WindTalker to sense victim’s ﬁnger clicks and the hand motion. Fig.12 shows the recovery accuracy of WindTalker in diﬀerent direction. It is interesting that WindTalker can achieve a high performance even the AP is deployed behind victims, which means that the proposed CSI based keystroke inference can work well even if the attacker is behind the user without visually seeing the clicking actions. This represents one of signiﬁcant merits which cannot be achieved by any previous work.

6. REAL-WORLD EXPERIMENT: MOBILE

PAYMENT PASSWORD INFERENCE TO- WARDS ALIPAY

6.1 System Setup

To demonstrate the practicality of the WindTalker, we perform an experimental evaluation of password inference on Alipay, a popular mobile payment platform on Both of Android and iOS system. Alipay is the largest mobile payments company in the world and has 450 million monthly active user including 270 million mobile payment users. As shown in Fig.13, we deploy a WindTalker system at a cafeteria-like environment and release an authentication-free WiFi. The AP (including Intel 5300 NIC and the antennas) is set up behind the counter, which makes it less likely to be detected visually. The victim is 1 meter away from our deployed WiFi devices. When we collect the data, one user walks pass by the victim but none of users walks between the victim and the AP.

To simulate the real-world attack scenarios, the recruited volunteers are required to access to this free WiFi access

1076

#### Page 10

Targe t

Hidden Devices Intel 5300 NIC

Antennas

1m

Figure 13: Real Case Scenario

points and perform the following three phases: 1) Online Training Phase: the volunteers are required to input some randomly generated numbers by following a similar way as Text Captchas. This phase is designed to collect the user’s input number and the corresponding CSI data to ﬁnish the data training. 2) Normal Use Phase: the volunteers perform the online browsing or use the applications as a normal user. 3) Mobile Payment Phase: when the users use the online shopping applications, it will be ended with the mobile payment. All of the online shopping and mobile payments are secured with HTTPS protocol. According to Alipay mobile payment policy, the mobile users must input the password to ﬁnish an mobile payment transactions. The goal of the attacker is to recover the mobile payment password of the volunteers.

6.2 Operations of WindTalker

After the volunteers connect to the authentication-free WiFi hotspot, WindTalker triggers ICMP based CSI Acquirement Module to collect the CSI data at the sampling rate of 800 packets/s. WindTalker records the timestamp per one hundred CSI data. Simultaneously, WindTalker utilizes Wireshark to capture and record WiFi traﬃc packets and their corresponding timestamp. During the realworld experiment, WindTalker collects WiFi traﬃc data and CSI data in the online phase. After collecting the data, WindTalker infers the user’s mobile payment password in the oﬄine phase.

6.3 Recognizing The Sensitive Input Windows:

To determine the sensitive input windows, WindTalker runs in a real-time fashion to collect the meta data (e.g., IP address) of the targeted sensitive mobile payment applications (e.g., Alipay). For example, in the experiment, Alipay applications will always route their data to the server of some speciﬁc IP address such as “110.75.xx.xx”. This IP address will be kept to be relatively stable for one or two weeks. With the traﬃc meta data, as shown in Fig.14(a), WindTalker obtains the rough start time and end point of Sensitive Input Window via searching packets whose destination is “110.75.xx.xx”. Then WindTalker begins to analyze the corresponding CSI data in that period of time.

6.4 CSI based Password Inference

Fig.14(b) shows the original 12th subcarrier CSI data in Sensitive Input Window. After data preprocessing, Fig.14(c) shows the ﬁrst three principal components of CSI data after PCA. It is found that in the real-world experiment that besides input payment password, victim may have other operations such as selecting credit card for payment in period

Start

End …

Time

Time

(a) Sensitive Input Windows Recognition Module

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 x 10 4 10

15

20

25

30

35

40

A CSI se q uence containin g a Sensitive :LQGRZ 

CSI Amplitude

(b) Original CSI

Sensitive Input Window

Keystroke Extraction

Keystroke Inference 7 7

9 9 3

1

(c) Keystroke Inference Module

Figure 14: WindTalker in Case Study

of time of Sensitive Input. In order to handle this situation, WindTalker only needs to ﬁnd a continuous keystroke of certain length. In our case, we are interested in continuous 6-bit password input since Alipay chooses 6-digit mobile payment password. Thus after keystroke extraction and recognition process, WindTalker is able to list possible password candidates according to probability. The top three password candidates in this case is 773919, 773619, 773916 while the actual password is 773919. We carry out the real-world experiment ten times, each time the password is diﬀerent. Our experiment results show that the attacker can successfully recover 2, 4, 7 and 9 passwords if allowing to try the password input for 5, 10, 50 and 100 times (or Top 5, 10, 50, and 100 candidates). This further demonstrates the practicality of the proposed attack in the practical environment.

7. DISCUSSIONS

7.1 Limitations

In this section, we discuss the main limitations of WindTalker. WindTalker’s high performance is achieved in an experiment environment. However, if we try to apply WindTalker in anytime and anyplace, we need to overcome the limitations as follows.

Hardware Limitations. In WindTalker, we use Intel 5300 NIC and Linux 802.11n CSI Tool[8]. In our experiments, it is observed that the system will crash when we perform ICMP based CSI data collection for iPhone or some version of android smart phones. This is because, according to the statement of the author of CSI Tool, it is very easy to crash when one Intel 5300 NIC works with other NICs (e.g., an iPhone). However, our implementation and evaluation on a wide range of smart phones (including XiaoMi phones, Nexus and Samsung phones) demonstrate the practicality of the proposed CSI based keystroke inference method. We will leave the issues of improving the compatibility of Intel 5300 NIC with a wider range of mobile devices to our future work.

1077

#### Page 11

Table 2: Recovery Rate and Loop Times Loop Times One Three Five Ten Recovery Rate 68.3% 73.3% 78.3% 81.7%

Fixed Typing Gesture. Currently, WindTalker can only work for the situation that the victim can only touch the screen with a relatively ﬁxed gesture and the phone needs to be placed in a relative stable environment (e.g., a table). In reality, the user may type in an ad-hoc way (e.g., the victim may hold and shake the phone, or even perform some other actions while typing). We argue that is a common problem for most of the side channel based keystroke inference schemes such as [2, 13, 16]. This problem can be partially circumvented by proﬁling the victim ahead or performing a targeted attack by applying the relevant movement model as pointed out by [13].

User Speciﬁc Training. Using WindTalker, the victim’s input can be recognized via the classiﬁers trained from the same user. In the real-world experiments, it is hard to adopt the classiﬁers trained by other people to infer the victim’s input. This is because diﬀerent people have diﬀerent ﬁnger coverage and clicking model. A large number of training data based on a wide range of training samples may overcome this limitation. In practice, the attackers have more choices to achieve the user speciﬁc training. For example, it can simply oﬀer the user free WiFi access and, as the return, the victim should ﬁnish the online training by clicking the designated numbers. It can also mimic a Text Captchas to require the victim to input the chosen numbers. We further analyze the impact of the number of training data on recovery rate in WindTalker. Table.2 shows the recovery rate increases with the training loop increases. Even if there is only one training sample for one keystroke, WindTalker can still achieve whole recovery rate of 68.3%.

7.2 Defending Strategies

One of the most straightforward defense strategies is to randomize the layouts of the PIN keypad, such that the attacker cannot recover the typed PIN number even if he can infer the keystroke positions on the touchscreen. As pointed out by [23], randomizing the keyboards is the eﬀective at the cost of the user experience since the user needs to ﬁnd every key on a random keyboard layout for every key typing.

A more practical defense strategy is preventing the collection of CSI data. For example, the user refuses to connect to free public WiFi or pays attention to the deployed WiFi devices nearby. Note that, to have the successful CSI based keystroke inference, the sender WiFi device should be deployed close enough to the victim (e.g., 30 cm as shown in [2]). To prevent the accurate CSI data collection, another strategy is obfuscating the CSI data by adding some randomized noises to CSI data. In particular, the user can intentionally change his typing gestures or clicking patterns, since ﬁnger coverage and click pattern are considered as two major factors that aﬀect CSI value for the keystroke. Further, since CSI reﬂects the change of multi-path propagation of WiFi signals, the users can take some actions to introduce the unexpected interferences to the CSI data. For example, the randomized human behaviors (e.g., human mobility) or wireless signals will reduce the successful chance of the adversary. Lastly, for the proposed ICMP based CSI collection

approach, CSI based typing inference requires collecting CSI data with a high frequency. Therefore, detecting and preventing a high-frequency ICMP ping represent a practical and ease of use countermeasure.

8. RELATED WORK

In this section, we review two domains of prior works that are tightly related to WindTalker.

8.1 Public free Wi-Fi with malicious behav-

iors Free Wi-Fi services provided by public hotspots are attractive to users in a mobile environment when their mobile devices have limited Cellular connection. Existing works [5, 6, 11, 21] have demonstrated it is feasible to deploy a malicious Wi-Fi hotspot in a public area. For example, an iPhone can turn itself into a Wi-Fi hotspot. If the iPhone user changes the session ID to “Starbucks Free Wi- Fi”, other people may connect their phones to the iPhone while wrongly believe they are using free WiFi services from a nearby Starbucks.

In our considered scenarios, attackers may make use of user’s trusts on pubic WiFi and lure the the users to connect their devices to a fake access point. Then, the attacker eavesdrops the WiFi traﬃc to identify the sensitive windows and selectively analyzes the CSI information to infer the keystroke information.

8.2 Keystroke Inference methods

Prior keystroke inference methods have been developed based on the information from various sensors and communication channels, such as motion, camera, acoustic signals, and WiFi signals.

Motion: Owusu et al. [16] presented an accelerometerbased keystroke inference method, which aims to recover six-character passwords on smartphones. Later, Liu et al. [13] applied a similar idea to the smartwatch scenario. Their objective is to track user’s hand movement over the keyboard using the accelerometer readings from the smartwatch, and the keystroke inference achieves 65% recognition accuracy.

Acoustic signals: Zhu et al. [25] presented a contextfree and geometry-based keystroke inference. They use the microphones at a smartphone to record keystrokes’ acoustic emanations. Liu et al. [12] further proposed a keystroke snooping system by exploiting the audio hardware to distinguish mm-level position diﬀerence. Their experiments showed the system can recover 94% of keystrokes.

Camera based: Yue et al. [23] introduces a camerabased keystroke inference using Google Glass or oﬀ-the-shelf webcam. This method can achieve a per-input success rate of over 90%. Shukla et al. [18] also presented a video-based attack relies on the spatio-temporal dynamics of the hands during typing. The paper can breaks an average of over 50% of the PINs. Sun et al. [19] use camera to record tablet backside motion and infer the victim’s typing content.

WiFi signal based: Using Wi-Fi signals to infer the keystroke recently draws a large research attention because it oﬀers device-free and non-invasion advantages. The channel state information (CSI) are obtained from the commercial Wi-Fi network interface cards. Many research works have demonstrated such ﬁne-grained information can be very eﬀective in detecting the ambient physical movement because it well captures the reﬂected multi-path WiFi signals.

1078

#### Page 12

Liu et al. [2] proposed a keystroke inference systems called WiKey, which uses the CSI waveform pattern generated by ﬁnger’s unique motion to distinguish keystrokes on a external keyboard. Compared with our work, WiKey works on the OKI keystroke inference model and it can not recognize the sensitive input windows. Zhang et al. [24] also presented WiPass, which can work in mobile device to detect the graphical unlock passwords.

9. CONCLUSION AND FUTURE WORK

In this paper, we have designed and evaluated a novel side-channel attack based on CSI which can infer victim’s input on smartphone via WiFi signals. Our evaluation shows that our attack can work well in recognizing the victim’s password on smart phones. Compared with the previous side channel based keystroke inference work, WindTalker neither deploys external devices close to the target device nor compromises the target device. It can even be launched behind the victim without the requirement of visually seeing the smart phone user’s input process, backside motion, or installing any malware on the tablet. Due to the limitation of Intel 5300 NIC, the current WindTalker cannot work for iOS smartphones, which will be a part of our future work. We will investigate how to further improve the inference accuracy of WindTalker under diﬀerent environments.

Acknowledgments

This work is supported by National Science Foundation of China (No. 61272444, U1401253, U1405251, 61411146001) and National Science Foundation (No. 1527144, No. 1553304, No. 1618893).

10. REFERENCES

[1] IEEE Std. 802.11n-2009: Enhancements for higher

throughput. http://www.ieee802.org, 2009. [2] Ali, K., Liu, A. X., Wang, W., and Shahzad, M.

Keystroke recognition using wiﬁ signals. In Proceedings of the 21st Annual International Conference on Mobile Computing and Networking (2015), ACM, pp. 90–102. [3] Balzarotti, D., Cova, M., and Vigna, G. Clearshot:

Eavesdropping on keyboard input from video. In Security and Privacy, 2008. SP 2008. IEEE Symposium on (2008), IEEE, pp. 170–183. [4] Benko, H., Wilson, A. D., and Baudisch, P. Precise

selection techniques for multi-touch screens. In Proceedings of the SIGCHI conference on Human Factors in computing systems (2006), ACM, pp. 1263–1272. [5] Cheng, N., Wang, X., Cheng, W., Mohapatra, P., and

Seneviratne, A. Characterizing privacy leakage of public wiﬁ networks for users on travel. In INFOCOM, 2013 Proceedings IEEE (2013), IEEE, pp. 2769–2777. [6] Fan, Y., Jiang, Y., Zhu, H., and Shen, X. S. An eﬃcient

privacy-preserving scheme against traﬃc analysis attacks in network coding. In INFOCOM 2009, IEEE (2009), IEEE, pp. 2213–2221. [7] Forlines, C., Wigdor, D., Shen, C., and Balakrishnan,

R. Direct-touch vs. mouse input for tabletop displays. In Proceedings of the SIGCHI conference on Human factors in computing systems (2007), ACM, pp. 647–656. [8] Halperin, D., Hu, W., Sheth, A., and Wetherall, D.

Tool release: gathering 802.11 n traces with channel state information. ACM SIGCOMM Computer Communication Review 41, 1 (2011), 53–53. [9] Hamed, K. H., and Rao, A. R. A modiﬁed mann-kendall

trend test for autocorrelated data. Journal of Hydrology 204, 1 (1998), 182–196.

[10] Holt,C.C.Forecasting seasonals and trends by

exponentially weighted moving averages. International journal of forecasting 20, 1 (2004), 5–10. [11] Konings, B., Bachmaier, C., Schaub, F., and Weber,

M. Device names in the wild: Investigating privacy risks of zero conﬁguration networking. In Mobile Data Management (MDM), 2013 IEEE 14th International Conference on (2013), vol. 2, IEEE, pp. 51–56. [12] Liu,J.,Wang,Y.,Kar,G.,Chen,Y.,Yang,J.,and

Gruteser, M. Snooping keystrokes with mm-level audio ranging on a single phone. In Proceedings of the 21st Annual International Conference on Mobile Computing and Networking (2015), ACM, pp. 142–154. [13] Liu, X., Zhou, Z., Diao, W., Li, Z., and Zhang, K. When

good becomes evil: Keystroke inference with smartwatch. In Proceedings of the 22nd ACM SIGSAC Conference on Computer and Communications Security (2015), ACM, pp. 1273–1285. [14] Lozowski, E., Charlton, R., Nguyen, C., and Wilson,

J. The use of cumulative monthly mean temperature anomalies in the analysis of local interannual climate variability. Journal of Climate 2, 9 (1989), 1059–1068. [15] Mar quardt, P., Verma, A., Carter, H., and Traynor, P. (sp) iphone: decoding vibrations from nearby keyboards using mobile phone accelerometers. In Proceedings of the 18th ACM conference on Computer and communications security (2011), ACM, pp. 551–562. [16] Owusu, E., Han, J., Das, S., Perrig, A., and Zhang, J.

Accessory: password inference using accelerometers on smartphones. In Proceedings of the Twelfth Workshop on Mobile Computing Systems & Applications (2012), pp. 1–6. [17] Sen, S., Lee, J., Kim, K.-H., and Congdon, P. Avoiding

multipath to revive inbuilding wiﬁ localization. In Proceeding of the 11th annual international conference on Mobile systems, applications, and services (2013), ACM, pp. 249–262. [18] Shukla, D., Kumar, R., Serwadda, A., and Phoha,

V. V. Beware, your hands reveal your secrets! In Proceedings of the 2014 ACM SIGSAC Conference on Computer and Communications Security (2014), ACM, pp. 904–917. [19] Sun, J., Jin, X., Chen, Y., Zhang, J., Zhang, R., and

Zhang, Y. Visible: Video-assisted keystroke inference from tablet backside motion. [20] Wang,F.,Cao,X.,Ren,X.,andIrani,P.Detecting and

leveraging ﬁnger orientation for interaction with direct-touch surfaces. In Proceedings of the 22nd annual ACM symposium on User interface software and technology (2009), ACM, pp. 23–32. [21] Xia, N., Song, H. H., Liao, Y., Iliofotou, M., Nucci,

A., Zhang, Z.-L., and Kuzmanovic, A. Mosaic: Quantifying privacy leakage in mobile networks. In ACM SIGCOMM Computer Communication Review (2013), vol. 43, ACM, pp. 279–290. [22] Xie,Y.,Li,Z.,andLi,M.Precise power delay proﬁling

with commodity wiﬁ. In Proceedings of the 21st Annual International Conference on Mobile Computing and Networking (New York, NY, USA, 2015), MobiCom ’15, ACM, pp. 53–64. [23] Yue, Q., Ling, Z., Fu, X., Liu, B., Ren, K., and Zhao,

W. Blind recognition of touched keys on mobile devices. In Proceedings of the 2014 ACM SIGSAC Conference on Computer and Communications Security (2014), ACM, pp. 1403–1414. [24] Zhang, J., Zheng, X., Tang, Z., Xing, T., Chen, X.,

Fang, D., Li, R., Gong, X., and Chen, F. Privacy leakage in mobile sensing: your unlock passwords can be leaked through wireless hotspot functionality. [25] Zhu, T., Ma, Q., Zhang, S., and Liu, Y. Context-free

attacks using keyboard acoustic emanations. In Proceedings of the 2014 ACM SIGSAC Conference on Computer and Communications Security (2014), ACM, pp. 453–464.

1079

---

Source: [Fermat's Library](https://fermatslibrary.com/) — https://fermatslibrary.com/s/when-csi-meets-public-wifi-inferring-your-mobile-phone-password-via-wifi-signals. More annotated papers: https://fermatslibrary.com/llms.txt
