Real Time Content Based Automated Privacy Policy QR Code Watermarking For User Uploaded Images in Cloud

IPASJ International Journal of Computer Science (IIJCS)
Web Site: http://www.ipasj.org/IIJCS/IIJCS.htm

A Publisher for Research Motivation ........ Email:editoriijcs@ipasj.org
Volume 5, Issue 10, October 2017 ISSN 2321-5992
Real Time Content Based Automated Privacy

Policy QR Code Watermarking for User
Uploaded Images in Cloud
Jayaramu H S1 and A K Gautham2
1
Research Scholar, Mewar university, Chittorgarh, Rajasthan, India
2
Research Supervisor, Mewar University, India
Principal, S D College of Eneering, Muzzaffarnagar, India.,
Abstract: It is been well known fact that the privacy has being one of the major concerns in data sharing over the cloud. Even
though every cloud service provider who allows user photo sharing (for example Flickr, Facebook, Instagram) will have their
own privacy settings by means of which user can determine who are authorized to view the photograph and share the uploaded
photographs. However, it is often found that several users are ignorant about these privacy settings which lead to mishandling
of sensitive digital photos. It is also seen often that due to wrong privacy settings, such photos end up appearing in the time
line of unintended users. Beside this, user tends to share the same photo over different platforms with different set of social
connectivity and privacy policy attached to them. Also there exists no automated tool for inferring privacy of the user shared
images and also a common protocol is missing by means of which independent providers can infer a privacy setting for a
photograph.In order to provide a solution for the aforementioned problem, we propose a novel privacy policy inference and
watermarking technique for user uploaded images. The contribution of this work is in uploading a manager tool that acts as
the bridge between user and cloud photo sharing platforms. The tool infers a set of privacy rules based on image metadata,
visual content and user's past privacy setting preferences and embeds the privacy information in the image using invisible
wavelet based watermarking. The non-encrypted watermarking information presented as a xml privacy chain is converted into
a QR code and embedded in the image which can be used by multiple platform providers as an universal privacy setting for the
photos.
Keywords: watermarking, Steganography, Cloud data
1. INTRODUCTION
A lot of photos will be shared specifically in the social networking sites like facebook and flicker. Now privacy is
one of the major concerns with the sharing of photos. Content based privacy which means that the images are of
different types, for example images containing kids, friends, various nature sites , etc and these images can be
identified through image content matching algorithm ; every image is composed of certain combination of pixels. Each
pixel represents certain texture, color and shape. By properly identifying the texture, color and shape of an image one
can find out category of an image i.e., whether it is a sports image or a nature image. So whenever one initially starts
accessing a site, the access ors behaviour is remembered by the system that means which type of photos that are shared
with friends, family and everyone.
One can also find out the category of the image from the text that is put with the image. For example, the user
while posting about new born kids photo will invariably write the details of the baby born so that immediately one can
understand that the photo is associated with the category called baby or its content is that of the baby of that user. Now
as the person keeps on sharing his photographs with some privacy setting at the beginning for example babys photo for
only family members, classmates photographs only for classmate members, sports photograph only for group associated
with sports, etc., So once the user starts seeking with privacy at the beginning, the algorithm remembers the users
privacy setting and tries to find out both textual content as well as the image content. For example, the text contents an
image that is associated with an image known as a metadata which is the description of the images.
1.1 Present State of Art: Presently all the social networking sites for example, twitter, face book, flicker,
Instagram have the image sharing facility out of them flicker and face book are being extensible used for sharing
image content. However, none of face book or flicker has got content based privacy matching site therefore in both face
book as well as twitter there is a manual setup of the privacy setting for every image that is being published. Therefore,
the major problem is that the security of the image data ultimately depends upon how careful the user is at the time of
Volume 5, Issue 10, October 2017 Page 72

publishing the photographs rather than the efficiency of the system which results in mishandling of several images in
several different ways. In order to prevent the protective images or the images which are not been desired to be used by
other entities to be leaked the proposed system automatically finds out the possible privacy setting for an image and
enforce it even if user forgets to use that privacy setting.
1.2 Contribution: The contribution of the proposed work is a client software (commonly referred as app) that runs on
a PC which is basically of flicker client. Flicker is a website where one can keep all sorts of images. The client will also
have a local repository of a set of rules inferred from past user's privacy settings during image uploading. These
inferences are linked to image features of the past uploaded images. The inferred privacy rule is first converted into a
xml privacy chain and is transformed into a QR code which is embedded in the images using wavelet watermarking.
This makes every image carry it's privacy settings across different platforms which can be leveraged by platform
providers to adopt the privacy settings for these independent photographs.
The major component of the work is as presented here:
1. The content extraction
2. Privacy estimation from content text metadata and image features
3. Privacy policy rule embedding in images.
The goal of this privacy app is to find out a privacy of the published image based of the content on the metadata of
the past images and by automatically enforcing the privacy policy for this new image based on the similar privacy
policy user used in the past. The embedded policy contains information about user's identity as well as set of rules for
the images. For simplicity of the concept we adopt two categories of popular privacy settings: Public and Private.
Because images goes through different scaling while being shared across social network sites. The ability of QR codes
to to be detected event with different scales and they being extremely noise tolerant makes the QR codes a natural
choice for the watermark seal of privacy.
2. RELATED WORK
Due to the using of social sites a huge data is being shared on this which is violating the privacy so for this a survey
has been done here and to prevent security a semantic annotated hidden Markova model is used to measure the
annotated photos similarity in the database. [1]
To keep security protection in community, images need to be protected through different settings Here a protection of
innovation prompt is used to share a data by the user this will fulfill the users end level goal. [2]
As the leaking of personal data within the friends or some group it is not satisfied to the user so to handle this type of
problem a survey has been taken place through which a review is given of different privacy settings for the user to
satisfy their level. [3]
Sharing of images on content sharing sites is causing a misusage of the users privacy. To save the people from
hackers it has been created a system called Adaptive Privacy Policy Prediction which is used here to generate a policy
on each image uploads. [4]
To provide security to the image and shared data Images which are to be published System automatically annotates
the image using hidden Markova mode and features are extracted. [5]
Uploading a photos in the content sharing sites may leads to a violating the privacy to avoid this they solved it by
providing a review through survey mainly to enhance the secure the personal information. [6]
To completely survey for a security and sharing image privately by the outline of new projection saving method for
labeling image on long range informal destination for the communication has been advised here. [7]
The answer will be known that how the similar policies obtained by automatically generation of the policy on each
uploaded photo so there is also a restricted to access on shared data and also how the effects on effectiveness is
approached on tagging system.[8]
3. PROPOSED WORK:
A user will first need to create an account in flicker. User will create some groups in the flicker then publishes
some photographs with appropriate metadata. Then for these different images, user should set the privacy policy
carefully. Once this is done , whenever the user submits an image, the features of the image both metadata i.e., the
text content as well as the content i.e., pixel, color, texture , the shape in the image will be automatically extracted
and set in the data base. Whenever user submits a new image for publishing, this image will be first classified based
on the features of the previously stored images So the metadata of the current image will be compared with that of

the previous image as well as features of the current image i.e., color, texture, shape will be compared with
previously shared images textual content. If the difference with current image is quite high with respect to the
previous images, the user will be asked to enter a new category for the currently submitted image. If the image
category matches with any of the earlier submitted images then the image privacy should be automatically set the
overall system enabling the automatic protection of the images such that unwanted entity should not be able to view
the images, for instance the group which is associated with sports should be able to view only sports images, the
group which is associated with political images should be able to see only political images , the family members
should be able to see only the images associated that is shared with the family. The derived Watermark is embedded
on the image using wavelet based invisible watermark.
4. METHODOLOGY:
User will input an image if already user had earlier put some image into flicker or something. Current images
features will be extracted features means color features will be extracted. They will be compared with the previous
images will get a metadata. Metadata means what were the tags of previous images, what were the security of previous
images this policy will be extracted from by comparing the image policy which is already stored by the user once this is
been done whenever this image is published this image will be published with the predicted policy.
Figure 1 The architecture Diagram of the proposed model
Figure 2: 0th level Image Classification DFD

Figure 3: 1st level Image Classification DFD

From the image, the features will be extracted, These features will be having a RGB standard deviation and mean
standard deviation through the K-Nn classifier it takes the nearest image and matches, then it will ask for the private or
public then query manipulation means giving a new image when you give a new image when you give a new image
again that new image will going to be again matching algorithm and again to the flicker.
The use case is user and the flicker whereas the proposed app is placed in between them. User gives auto tags, takes tag
and adds privacy to this. User first needs to authenticate before uploading images and then search for the images
globally. In flicker, first code is generated then it will be verified and result is searched.
Figure 4: First and second stages of the case diagram of the model

Figure 5: Sequence diagram of the Model

In the beginning, the user will be giving the training our objective is to build an app. In this app user will keep some
images simply .Net based app user will give some image to this app will give privacy setting whom he wants to see
some text and tag the app will going to extract the color feature and then store those features along with the tag along
with the privacy in a local database and put the same thing to twitter. So many of the images as to be put in this ways
Once user does that next time he goes to the testing process. Whenever he gives an image from that image app extracts
the feature this gives the feature to the classifier. classifier also retrieves the earlier store features from the data base it
find outs the closest image then it finds out the privacy of the closest image then the current image that user wants to
put into database will be put into flicker the extracted privacy policy.
Watermarking
Firstly user's Identity and the privacy settings are put in a xml chain and is converted to a QR code.
For example the privacy xml of following settings
<name>HS Jayaramu</name><privacy>public</privacy>
is converted into the QR code which is shown in figure 8
Figure 6: Interaction Phase of Collaboration Figure 7: Activity Diagram

Figure 8: Sample of Generated QR Code

Wavelet based watermarking technique is adopted for embedding this information in the source image. As QR codes
are monochrome and shared source images are often color images, we extract the red channel of the image for
watermarking the code.
Figure 9: Overall block diagram is depicted below.

The proposed technique is based on singular value decomposition of Low pass filter channel of the wavelet transformed
image and applying the source pixel directly to non zero dc component of the channel. The combined components are
scaled back to original image scale through inverse transform. As QR codes themselves are binary representation,
wavelet transform is not needed on the watermarking image and the temporal data is directly embedded into the source
image red channel's LL component. This reduces the complexity of the overall process.
Further, as images loose their embedded information upon JPEG encoding, the watermarked images are converted to
PNG images before uploading. Though PNG has more spatial complexity in comparison to PNG images and requires
more bandwidth for both uploading as well as downloading. This is the only drawback of the method. But considering
the number of privacy breach related online issues, this is a very small tradeoff to protect user's privacy.

5. RESULT ANALYSIS
The algorithm was tested for different samples which had different security levels in the users flicker account. The
result is as below which shows almost 87% of accuracy as it is evident from the table below. It may be observed that
there are less number of false outputs.
Table 1: Analysis of obtained results.
Figure 10: Showing he Decoding accuracy levels for both LSB and Wavelet based algorithms
23
Present
22.5 Proposed
22
21.5
21
PSNR
20.5
20
19.5
19
18.5
25 30 35 40 45 50
Noise Percentage
Figure 11: PSNR performance upon addition of Salt and Pepper Noise

Figure 12: PSNR performance upon addition of Salt and Gaussian Noise
One can apply salt and peepers noise from the scale of 1 to 8 (10% noise to 80%) and analyze the post decoding
information extraction success for the 50 image custom dataset we created for the privacy testing analysis. The results
are as presented above. It can be clearly seen that proposed wavelet based information embedding yields better
accuracy.
The PSNR performance of the noise susceptibility of the proposed wavelet based technique has been analysed with that
of LSB based encoding technique after decoding the image.
6. CONCLUSION
With the popularity of the social networking sites every day millions of photographs have been shared in the social
networking site. This increases the risk of misusage of the photographs been shared. Many a times, user forgets to set
appropriate security and privacy setting for the images that are shared across the social site. In this paper, a novel
mechanism has been proposed to guide the user to automate the process of ensuring privacy setting for the images. Also
a wavelet based watermarking technique to embed the privacy policy through a corresponding QR code has been
proposed. The proposed technique first learns from already shared images by the user about the pattern and then
classifies any new image that user intends to share in social site of flicker as either private or public. Results show that
the proposed system can predict the privacy setting of the images with an accuracy of over 80% therefore this can be
used in a large variety of application and domain including facebook, twitter, Google plus etc...
Reference
[1] K.Mayuri, V.Divyavani, Y.SubbaRayudu Enhanced Privacy Access Inference for user uploaded images for Images
Sharing sites in web. International Journal of Computer Science, October 2015.
[2]K.Ramya Krishna, Smitha Rani Sahu, The /security method provocation for user assigned images on content
sharing website. International Journal of Computer Science and Technology, IJCST VOL 7, Issue 1, Jan-March
2016.
[3]Sangeetha.J, A Survey on the Privacy Settings of User Data and Images on Content Sharing Sites. International
Journal of Innovative Research in Computer and Communication Engineering , March. Volume 3, Issue 3, March
2015.
[4] Anna Cinzia Squicciarini, Dan Lin, Smitha Sundareswaran and Joshua Wede, Privacy Policy Inference of User-
Uploaded Images on Content Sharing Sites, IEEE Transactions on Knowledge and Data Engineering, Vol. 27,
NO. 1, January 2015.
[5]Sangeetha.J, Kavitha.R An improved privacy policy inference over the socially shared images with automated
annotation process, International Journal of Computer Science and Information Technologies, Volume.6 (3),
2015.
[6]Ashita A review of various privacy policy approaches to improve security in social networking communities.
International Journal of Advanced Research in Computer Engineering & Technology (IJARCET) Volume 4 Issue
12, December 2015

[7]Thumpilli Buthcinaida, Ch.K. Reepes kumar, Security approach induction of client transferred picture on content
sharing websites, International Journal of Research in Computer and Communication Technology, Volume 4,
Issue 12, Dec- 2015.
[8] K.Soniya lakshmi, V.subhatra, D.Gowdhami, Privacy policy inference model using enhanced parent control
algorithm, International Journal of Modern Trends in Engineering and Research Vol 02, Issue 11, November
2015.
AUTHOR
H S Jayaramu received the BE degree from Malnad College of Engineering, Hassan. MTech MTech
degree from SJCE Mysore, MS degree from Birla Institute of Technology and Science BITS, Pilani,
Rajasthan. He has got more than 35 years of teaching experience in various institutions and he has over
13 research publications in National and international conferences and Journals. Currently he is
working as Professor in the Dept. of Telecommunication Engineering, Sri Siddhartha Institute of
Technology, Tumkur. His research interests include Signal processing, image processing, Logic design,
and Steganography.
Dr Arvind K Goutam received his M Tech degree in Electronics & Communication Engineering
from Rajastan VidhyaPeet & another M Tech degree in Instrumentation Engineering from R E C
Kulakshetra and PhD from Meerut University. He has got 20 years of teaching experience and has 40
research publications in National and International conferences. Currently he is working as Principal,
S D College of Engineering, Muzzaffarnagar, Uttar Pradesh. His research interests include image
processing

Real Time Content Based Automated Privacy Policy QR Code Watermarking For User Uploaded Images in Cloud

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Real Time Content Based Automated Privacy Policy QR Code Watermarking For User Uploaded Images in Cloud

Diunggah oleh

Hak Cipta:

Format Tersedia

IPASJ International Journal of Computer Science (IIJCS)

Web Site: http://www.ipasj.org/IIJCS/IIJCS.htm

Real Time Content Based Automated Privacy

Volume 5, Issue 10, October 2017 Page 72

Volume 5, Issue 10, October 2017 Page 73

Figure 1 The architecture Diagram of the proposed model

Figure 2: 0th level Image Classification DFD

Volume 5, Issue 10, October 2017 Page 74

Figure 3: 1st level Image Classification DFD

Volume 5, Issue 10, October 2017 Page 75

Figure 5: Sequence diagram of the Model

Figure 6: Interaction Phase of Collaboration Figure 7: Activity Diagram

Volume 5, Issue 10, October 2017 Page 76

Figure 8: Sample of Generated QR Code

Figure 9: Overall block diagram is depicted below.

Volume 5, Issue 10, October 2017 Page 77

Volume 5, Issue 10, October 2017 Page 78

Volume 5, Issue 10, October 2017 Page 79

Volume 5, Issue 10, October 2017 Page 80

Anda mungkin juga menyukai