Expert Answer:GRAD 695 Harrisburg Anonymizing Voice Input in Mob

  

Solved by verified expert:Submit your introduction here. Your introduction must contain the following: Background, Problem Statement, Purpose Statement, Research Questions and Objectives, Significance of Study, Definition of Terms, Theoretical Framework, Hypothesis(es), Ethical Considerations, Delimitations, and Limitations. Papers that do not have these components as second level headings will lose points. You must also write in the APA style.
vermamunish_grad695_assignment1.docx

vermamunish_grad695_assignment2.docx

Don't use plagiarized sources. Get Your Custom Essay on
Expert Answer:GRAD 695 Harrisburg Anonymizing Voice Input in Mob
Just from $10/Page
Order Essay

Unformatted Attachment Preview

Running Head: INTRODUCTION
1
Anonymizing Voice Input in Mobile Devices
Munish Verma
Harrisburg University of Science and Technology
Assignment Submitted to Dr. Stanley Nwoji
In Fulfillment of GRAD 695
February 5, 2019
ANONYMIZING VOICE INPUT IN MOBILE DEVICES
2
Table of Content
Abstract …………………………………………………………………………………………………………………………. 3
Introduction ……………………………………………………………………………………………………………………. 4
Existing Risks ………………………………………………………………………………………………………………… 5
Privacy Risks ……………………………………………………………………………………………………………… 5
Security Risks …………………………………………………………………………………………………………….. 5
Proposed Solution …………………………………………………………………………………………………………… 6
SMART Criteria……………………………………………………………………………………………………………… 7
References ……………………………………………………………………………………………………………………… 8
ANONYMIZING VOICE INPUT IN MOBILE DEVICES
3
Abstract
Human-computer interaction (HCI) refers to the study of people’s interaction with
computers and the extent to which computers are or aren’t developed for interacting with humans
successfully. Historically, excluding a few, most computer system developers failed to pay
enough attention to ease of use of computers. Arguably, even today some computer users would
say that not enough attention is paid by computer makers in ensuring their products are user
friendly. Computer system developers may however defend themselves by illustrating the
extreme complexity of designing and making computer products and that the demand for
services rendered by computers has always beaten the demand for ease of use. Mobile devices
nowadays use an input class that offers users access to features such as touchscreen,
geographical input and accelerometer and even voice input. Through voice input, user experience
has greatly improved by allowing them to operate their devices hands free. A technology key to
voice input is speech recognition, which often gets outsourced to the cloud for optimal
performance. Outsourcing to the cloud however, comes with its own risks since the privacy of a
user can easily be compromised by matching a voice to a specific user, using speech recognition
to learn their sensitive input content and further profiling users based on such content.
This paper aims to reveal possible solution to the security concern earlier mentioned,
through an intermediate between users and the cloud, which creates user anonymity by sanitizing
the voice data of users before they get sent to the cloud for speech recognition. The technique
adopts a two-phase sanitization model i.e. an attack resistant voice conversion mechanism and
evolution-based keyword substitution. The two phases are all performed in a user’s mobile
device while still ensuring the cloud-supported speech recognition service is accurate and usable.
I estimate that this can minimize the probability of a user’s voice being identified by up to 85%.
ANONYMIZING VOICE INPUT IN MOBILE DEVICES
Introduction
When you take a look at hands-free communication, the applications of voice
input in mobile devices are broad especially in areas such as voice search e.g. Microsoft
Bing and Google, keyboard applications e.g. Microsoft Sougou, Google, Swift key and in
artificial intelligence personal assistants e.g. Google assistant, Apple’s Siri and Amazon
Echo. Voice input tremendously improves user experience and lives by freeing users’
hands and attention from time-consuming typing on the usually small screen of mobile
devices and is also the major means through which visually impaired people achieve
human-computer interaction. Speech-to-text conversion, also known as speech
recognition, is a key technology behind successful voice input. Through it, users’ vocal
input in form of spoken language gets recognized then translated into text. However, due
to limited resources on mobile devices, speech recognition usually gets outsourced to
cloud servers for optimal efficiency and accuracy.
Most voice input service providers collect users’ speech records since once the
voice datasets get correlated to the speakers’ identities, plenty of private information that
describes the individual can be revealed. Such Personally Identifiable Information should
never get disclosed on the cloud. At the moment, it is achievable by leveraging on
anonymous networks like Tor however, due to inefficient voice sanitization, the voice
records can still get de-anonymized using linkage attacks, which are possible due to the
varying voice biometrics of different individuals.
The paper therefore gives a glimpse into a way of protecting the speech data
privacy of voice input users without compromising their user experience. It also shows
how this particular research topic is SMART.
4
ANONYMIZING VOICE INPUT IN MOBILE DEVICES
5
Existing Risks
Privacy Risks
Broadly speaking, the two privacy risks that cloud-based voice input services face are;

The cloud or third parties responsible for obtaining voice data from the cloud linking a user’s
speech records to the actual individual by leveraging on the speech recognition technology.
In the event that the cloud manages to collect several speech samples of a specific user from
various sources e.g. YouTube, then train a voice model of the user, it may be able to identify
the user’s records from the speech database leading to identity privacy breach.

The cloud could also conduct an analysis of the content of a user’s speeches to gather
detailed private information. Content privacy breach can be conducted if the cloud uses
natural language processing to gather information from a user’s voice command and voice
search history or even emails and text messages that were typed using voice input.
Through this, the cloud can accurately determine a user’s personal preferences, habits,
demographic characteristics, schedules and communication.
Security Risks
However, besides the mentioned privacy risks, there could also arise some security risks
such as spoofing attacks to speech authentication systems. The use of voice for biometric
information has risen in modern authentication systems e.g. unlocking smart phones, gaining
access to apps e.g. WeChat and activating services like OK Google or Hey Siri. The cloud can
repeatedly collect a user’s voice samples to accurately create a voice print for synthesizing other
speeches identical to the user’s voice. Such can be used to spoof voice authentication systems and
gain unauthorized access to a user’s private information or even worse, be used to replicate the
user’s voice and create recordings for blackmailing the user.
ANONYMIZING VOICE INPUT IN MOBILE DEVICES
Proposed Solution
To mitigate the risks earlier mentioned, privacy preservation can be broadly achieved
through; securing a user’s identity privacy and preserving the content privacy of their
speeches. This however presents a fourfold challenge;

First, a user’s speech data should be sanitized without compromising the efficiency and
accuracy of speech-to-text conversion.

Second, protecting speech content privacy can be difficult since it requires the making of
certain considerations i.e. which information do users consider private? How can the private
content be located within a user’s speech? and how can the private content be hidden once
identified, without compromising user experience?

Third, speech in its unstructured data form is challenging to process in a privacy preserving
manner since conventional multi-party computation and cryptographic tools e.g. searchable
encryption can’t be relied on.

Fourth, efficient, real-time speech sanitization is difficult to perform considering a mobile
device’s limited resources.
To mitigate the risks aforementioned and ensure the cloud can only access sanitized
speech data, I propose the introduction of an intermediary application, which connects
mobile devices to the cloud by acting as a third-party voice input application. Audio signals
received on a mobile device get processed by the sanitizer which then transmits sanitized
audio to the cloud. The sanitization is done by; disguising a speaker’s identity via voice
conversion technology to modify the user’s voice and preventing the disclosure of sensitive
content to the cloud by perturbing the content of voice input through keyword substitution.
6
ANONYMIZING VOICE INPUT IN MOBILE DEVICES
7
The two methods complement each other to ensure both the voice input app and the cloud don’t
learn a user’s private data and their identity.
SMART Criteria
Smart is an acronym that stands for; Specific, Measurable, Attainable, Relevant and Time-Based.
The chosen research topic conforms to the criteria because;

Specific- It focusses on the specific issue of the minimal anonymity that voice input users
face and how it can be possibly addressed.

Measurable- The success of the voice input anonymity can be gauged by the extent to which
users’ identities and private information is not disclosed to the voice input apps, the cloud or
unauthorized third parties, all without compromising user experience.

Attainable- Voice input platforms already exist and several privacy-preserving systems for
voice input data have been and are being developed e.g. voice mask application.

Relevant- There exists an issue of user anonymity and inherent security risks they are
exposed to by using voice input techniques, which can be addressed through the proposed
solution.

Time-based- In the next four years, there will be well functioning privacy-preserving
systems for voice input data to address its shortcomings.
ANONYMIZING VOICE INPUT IN MOBILE DEVICES
8
References
Cardinal, D. (2018, March 5). From Punchcards to Siri: The history (and future) of input
devices. Retrieved from ExtremeTech: https://www.extremetech.com/computing/98287from-punchcards-to-ipads-the-history-of-input-device/8
Du, H. &. (2019, January 13). Voice Mask: Anonymize and Sanitize Voice Input on Mobile
Devices. Retrieved from Academia:
https://www.academia.edu/35863738/VoiceMask_Anonymize_and_Sanitize_Voice_Inpu
t_on_Mobile_Devices
Here Are Some Business Goals That Follow the SMART Criteria. (2019). Retrieved from
https://www.thebalancesmb.com/smart-goal-examples-2951827
Lyndsay. (2018, March 5). Human Computer Interaction. Retrieved from Search Software
Quality: http://searchsoftwarequality.techtarget.com/definition/HCI-human-computerinteraction
Running Head: INTRODUCTION
1
Anonymizing Voice Input in Mobile Devices – Literature View
Munish Verma
Harrisburg University of Science and Technology
Assignment Submitted to Dr. Stanley Nwoji
In Fulfillment of GRAD 695
February 17, 2019
ANONYMIZING VOICE INPUT IN MOBILE DEVICES
2
ANONYMIZING VOICE INPUT IN MOBILE DEVICES
3
Table of Content
Anonymizing Voice Input in Mobile Devices …………………………………………………………………….. 4
Privacy and Security Risks …………………………………………………………………………………………… 4
Anonymizing Voice Input ……………………………………………………………………………………………. 5
Conclusion …………………………………………………………………………………………………………………. 7
References ……………………………………………………………………………………………………………………… 8
ANONYMIZING VOICE INPUT IN MOBILE DEVICES
Anonymizing Voice Input in Mobile Devices
Growth and advancement in technology has seen computer system developers
come up with more and more features which are aimed at improving user experience.
These features range from touchscreens, voice input, geographical input and so forth.
Voice input has specifically transformed user experience allowing its users to
comfortably operate their computer devices without necessarily using their hands. As
(Hawley, Cunningham, & Green, 2013) assert the implementation and adaptation of
voice input has even enabled people with severe speech impairment to utilize this
technology. Although the benefits have been great, the advancement has also come with
the risks of privacy and security for the users.
Privacy and Security Risks
Scholars (Qian, Du, Che, Wang, & Deng, 2017) assert that the application of voice input
in mobile devices has been every instrumental in improving user experience. However, they
argue that the technology behind voice input is speech recognition which could be easily
compromised since it’s normally outsourced to the cloud. The scholars argue that the privacy of
voice input users is at risk of compromise since the identities of the users could be unveiled and
their private sensitive information accessed. To these scholar’s privacy issues that have to be
contended with in the use of voice input are the user’s identity together with their content. This
they indicate could be made possible by profiling the users and using their inputted content to
know who they are.
The main reason why outsourced cloud services and the whole system of voice-input is
susceptible to compromise is because the parties who are contacted to serve as the providers of
voice input service keep records of their users’ speech. The danger lies in the fact that if the
4
ANONYMIZING VOICE INPUT IN MOBILE DEVICES
5
identities of the individual can be correlated with this voice data sets and personal identifiable
information could be revealed comprising the privacy and security of the individuals. (De Leon,
Pucher, Yamagishi, Hernaez, & Saratxaga, 2012) asserts that another security danger that lies
with the current system is that when enough sample information is collected in the cloud, it can
be exploited to create accurate and identical user voice print which is dangerous. Scholars (Wu &
Li., 2014) further indicate that with this voice datasets which have been collected, exploited and
synthesized to replicate the users and their mannerism, the security of the users can be
compromised. This synthesized speech can also be used to spoof various systems which require
voice authentication to try and get access to the private resources of the users
(Wu & Li., 2014) also assert that these voices can also be used to conduct illegal
operations such as fraud or come up with illegal recording that can be used to blackmail the
voice-input users for financial or other actions or frame them for a crime. In their study (Alepis
& Patsakis, 2017) sought to examine the privacy and security of voice assistants. They asserted
that while these systems were very helpful and greatly improved user experience, they opened up
a risk broad way in the security and privacy of their users more so in the era of “Internet of
Thing”. Their argument was based on the fact in the recent past the capabilities of research
assistants have been increasing and since they rely on voice input could be easily compromised.
Anonymizing Voice Input
The security and privacy of personal identifiable information of voice-input users
continues to be dangerously compromised. As (Gonzalex-Dominguez, Eustis, Moreno, Beaufays,
& Moreno, 2015) assert, millions of people globally use automatic speech recognition systems in
their day to day operations ranging from do searches, dictate messages, facilitate the input of
ANONYMIZING VOICE INPUT IN MOBILE DEVICES
6
data and control devices. This creates a large data base of voice-input which can be exploited to
compromise their security and privacy.
In order to safeguard the security and privacy of voice input users, this study proposes
that their data voices should be made anonymous. This is a position that has also been proposed
by scholars (Qian, Du, Che, Wang, & Deng, 2017)who argues that in order to safeguard the
security and the privacy of voice input users, a Voice Mask should be incorporated to serve as a
mediator between the users and the cloud in order to sanitize their voices and eliminate the
chances of compromise. The scholars assert that the way that the Voice Mask would operate was
by having a voice conversion mechanism and one which is resistant to attacks. They also
asserted that the mediator would also have a technique that substitutes the keywords of the voice
content providing additional protection. Employing these techniques, the scholars tested their
theories and determined that they successfully reduced identity compromise through voice input
recognition by 84% for their test group of 50 participants.
In order to convert the voice of voice input users and a prevent compromising their
identities; there are a number of technologies which can be used. (Sundermann & Ney, 2003)
identified frequency warping as one of the most common paradigms for voice conversion. This
technique which relies on vocal tract length normalization has received considerable interest and
study. This study therefore recommends them for anonymizing the voice of the users.
On the other hand, the other privacy issue is the content of the voice-input users’
message. In order to ensure its privacy, this study recommends the substitution of keywords.
Scholars (Chen, Parada, & Heigold, 2014) and (Zhang & Glass, 2009) have given considerable
attention to keyword spotting applications which identifies the key words from utterances made
directly. These scholars asserted that these applications are used to monitor keywords, to index
ANONYMIZING VOICE INPUT IN MOBILE DEVICES
7
document and for dialogue systems which are automated. In order to protect the privacy of the
voice input user content, this study proposes that these applications can be used to identify and
replace keywords.
The issue of preserving the privacy and security of voice-input data continues to emerge
as one that requires attention and considerable in put in order to address. (Qian, Du, Che, Wang,
& Deng, 2017) in their study sought to quantify the risks posed by the publication of speech data
and recommended that countermeasures should be taken to secure its privacy. (Erlingsson &
Korolova, 2015) asserted that the best way to ensure privacy was to employ an ordinal response
for randomized keyword aggregation. Along with other scholars who have proposed the same
this study proposes that an algorithm which performs randomized keyword aggregation should
be employed to ensure the privacy of voice input.
Conclusion
The use of voice input in various technological devices has served to improve the
experience of the users freeing their hands and enabling them to get results much faster.
However, the technology has also opened a can of worms on the issue of privacy and security of
the users since the stored data sets are potentially subject to compromise and abuse. Two areas of
security risk have been identified namely user identity compromise and data content
compromise. Therefore, in order to resolve the challenge and close any loopholes for
compromise, this study proposes anonymizing the voice input through voice conversion
mechanism and the substitution of keywords. It further proposes that this anonymizing should be
done by a mediator who connects the vice input user and the cloud. These steps are projected to
safeguard the privacy and security of voice input …
Purchase answer to see full
attachment

Place your order
(550 words)

Approximate price: $22

Calculate the price of your order

550 words
We'll send you the first draft for approval by September 11, 2018 at 10:52 AM
Total price:
$26
The price is based on these factors:
Academic level
Number of pages
Urgency
Basic features
  • Free title page and bibliography
  • Unlimited revisions
  • Plagiarism-free guarantee
  • Money-back guarantee
  • 24/7 support
On-demand options
  • Writer’s samples
  • Part-by-part delivery
  • Overnight delivery
  • Copies of used sources
  • Expert Proofreading
Paper format
  • 275 words per page
  • 12 pt Arial/Times New Roman
  • Double line spacing
  • Any citation style (APA, MLA, Chicago/Turabian, Harvard)

Our guarantees

Delivering a high-quality product at a reasonable price is not enough anymore.
That’s why we have developed 5 beneficial guarantees that will make your experience with our service enjoyable, easy, and safe.

Money-back guarantee

You have to be 100% sure of the quality of your product to give a money-back guarantee. This describes us perfectly. Make sure that this guarantee is totally transparent.

Read more

Zero-plagiarism guarantee

Each paper is composed from scratch, according to your instructions. It is then checked by our plagiarism-detection software. There is no gap where plagiarism could squeeze in.

Read more

Free-revision policy

Thanks to our free revisions, there is no way for you to be unsatisfied. We will work on your paper until you are completely happy with the result.

Read more

Privacy policy

Your email is safe, as we store it according to international data protection rules. Your bank details are secure, as we use only reliable payment systems.

Read more

Fair-cooperation guarantee

By sending us your money, you buy the service we provide. Check out our terms and conditions if you prefer business talks to be laid out in official language.

Read more

Order your essay today and save 30% with the discount code ESSAYSHELP