Secure and Trust-aware Computing for Drone-assisted IoT-Sensor Networks - Research Article

International Journal of Distributed

Sensor Networks
2020, Vol. 16(3)
Partial image encryption using Ó The Author(s) 2020
DOI: 10.1177/1550147720914779
format-preserving encryption in image journals.sagepub.com/home/dsn

processing systems for Internet of

things environment

Wonyoung Jang1 and Sun-Young Lee2

Abstract
Concomitant with advances in technology, the number of systems and devices that utilize image data has increased.
Nowadays, image processing devices incorporated into systems, such as the Internet of things, drones, and closed-circuit
television, can collect images of people and automatically share them with networks. Consequently, the threat of inva-
sion of privacy by image leakage has increased exponentially. However, traditional image-security methods, such as pri-
vacy masking and image encryption, have several disadvantages, including storage space wastage associated with data
padding, inability to decode, inability to recognize images without decoding, and exposure of private information after
decryption. This article proposes a method for partially encrypting private information in images using FF1 and FF3-1.
The proposed method encrypts private information without increasing the data size, solving the problem of wasted stor-
age space. Furthermore, using the proposed method, specific sections of encrypted images can be decrypted and recog-
nized before decryption of the entire information, which addresses the problems besetting traditional privacy masking
and image encryption methods. The results of histogram analysis, correlation analysis, number of pixels change rate, uni-
fied average change intensity, information entropy analysis, and NIST SP 800-22 verify the security and overall efficacy of
the proposed method.

Keywords
Partial image encryption, format-preserving encryption, FF1, FF3-1, IoT

Date received: 7 November 2019; accepted: 3 March 2020

Handling Editor: Dush Nalin K Jayakody

Introduction and protected by encryption based on a security algo-

rithm.8–20
With continuing advancements in the development of Privacy masking and image encryption are com-
image processing equipment, image data, such as monly applied as image data security measures. Privacy
images of people, are increasingly being collected and
used in various fields. Image processing equipments are 1
Department of Computer Science & Engineering, Soonchunhyang
also increasingly being connected to network environ- University, Asan, South Korea
ments. Images are intuitive and sensitive, which means 2
Department of Information Security Engineering, Soonchunhyang
that without proper privacy protection, data leaks, University, Asan, South Korea
such as privacy breaches, and confidential information
Corresponding author:
leaks can occur.1–7 As such threats increase in severity Sun-Young Lee, Department of Information Security Engineering,
and number, concerns about privacy are growing, spur- Soonchunhyang University, Asan 31538, Chungnam, South Korea.
ring the requirement for image information to be stored Email: sunlee@sch.ac.kr

Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(https://creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work
without further permission provided the original work is attributed as specified on the SAGE and Open Access pages
(https://us.sagepub.com/en-us/nam/open-access-at-sage).
2 International Journal of Distributed Sensor Networks

masking involves methods such as blurring, mosaic privacy masking methods, image encryption methods,
masking, removal, and deformation. Blurring of a spe- and format-preserving encryption. Section
cific area in an image makes it difficult to recognize ‘‘Contribution’’ introduces the advantages of the pro-
objects. Mosaic masking is a technique that blurs a spe- posed method. Section ‘‘Proposed method’’ describes
cific area of a picture by weighting it in a tile form. the proposed new method for encrypting parts of
Removal and deformation involve deleting data in a images using format-preserving encryption. Section
specific area or modifying data using arbitrary values. ‘‘Experiment evaluation’’ compares the encryption
These privacy masking techniques all involve overwrit- results with traditional image security. Sections
ing data in the original image. Although they are fast ‘‘Performance’’ and ‘‘Optimal block size’’ measure the
and efficient because of the low performance load encryption/decryption speed and derives the optimal
involved, they make it impossible to recover the origi- encryption block size, respectively. Section ‘‘Security
nal data.21 This results in the problem of such methods evaluation’’ presents the results of security analyses and
only being applicable with images that do not need to other comparisons of the proposed method with exist-
be recovered, unless the original image is stored sepa- ing image-security techniques. Conclusions are pre-
rately. There are two methods of encrypting image sented in section ‘‘Conclusion.’’
data: (1) full encryption, which encrypts the entire
image using the likes of block-cipher-based and chaos-
map-based encryption, and (2) partial encryption, Related work
which selectively encrypts only the sensitive parts of an
image. Because of the nature of image file structures, it Closed-circuit television camera system model
is difficult to encrypt only a portion of an image using A closed-circuit television (CCTV) camera system takes
image full encryption. Therefore, encrypted images can- a picture of an image and stores it as data in a manner
not be recognized by users because all of the data must that a user can see. In a CCTV system, a CCTV camera
be encrypted—not only the data that would result in is connected to an image recorder. A digital video
privacy violations if revealed. Private information in an recorder (DVR) processes the codec algorithm and
image is defined as a visible face, part of the body, or analog image that fit the image sent from various types
vehicle registration number that could be used to iden- of CCTVs into the type and form that fits the intended
tify an individual. When using a block cipher, which is purpose. The video is stored on a hard disk in the
typically used to encrypt text data for image encryp- DVR machine and is transmitted to a video storage ser-
tion, the size of the encrypted image can be increased ver using Secure Sockets Layer (SSL)-encrypted com-
by padding. In particular, for an image, when the munication. The video storage server encrypts and
amount of data increases due to padding in every stores all the video sent from the DVR to protect pri-
frame, storage space may be wasted. In this article, we vate information. Control room operators and external
propose a partial image encryption method using users can see the images saved in the data storage ser-
format-preserving encryption that can protect informa- ver with appropriate user authentication. Figure 1
tion requiring privacy and utilize images while over- shows a schematic illustration of the CCTV camera
coming these problems. Format-preserving encryption system model.
is an algorithm that is suitable for encryption of data
that have a fixed length or format and characteristics
that do not increase the length of the cryptogram. Privacy masking techniques
Therefore, it is possible to encrypt only the private
information in an image, and because the data size does Privacy masking techniques, such as blurring, mosaic
not increase, the encrypted data can be stored effec- masking, removal, and deformation, protect sensitive
tively without wasting storage space. The proposed or personal information by changing some data in an
method does not result in an increase in the amount of image so that the original information cannot be seen.
data due to the application of padding when encrypting When these methods are used, it is difficult to recover
only part of an image. In addition, the proposed the original image. Blurring and mosaic masking tech-
method permits encrypted images to be recovered, niques blur the original image so that it cannot be prop-
addressing a problem associated with privacy masking. erly recognized. Removal and deformation methods
Furthermore, it allows encrypted images to be recog- delete original image pixel values or overwrite original
nized by encrypting only private information. This arti- image pixel values with different values using arbitrary
cle describes the proposed method and presents the values. Privacy masking methods do not permit recov-
results of comparisons of the method with existing ery of the original image because the original image
video security methods. The remainder of this article is data are modified. Another disadvantage of such meth-
organized as follows. Section ‘‘Related work’’ gives an ods is that the original image must be inferred, although
overview of image processing systems, traditional the exact original is unknown.
Jang and Lee 3

Figure 1. CCTV camera system.

Image encryption information for encryption: plain text, a key, and a

There are two ways to protect an image from eavesdrop- tweak. It does not matter if the tweak is public. Even if
ping and leakage: full encryption and partial encryption. the key is the same as the plain text, it is safe from code-
book attack because the cipher text changes every time
Full image encryption—encryption of the entire image—
the tweak changes. The method proposed in this article
is used when it is necessary to perform encryption effi-
uses FF1 and FF3-1, standard format-preserving
ciently and quickly. Symmetric key block-cipher-based
encryption methods. Figure 2 shows the Feistel struc-
or chaotic-map-based image encryption methods, as well
ture of FF1 and FF3-1.
as convolution operation-based and hyper-chaos-based
FF1 and FF3-1 divide plain text into components A0
encryption, are currently being studied.22–30 Partial
and B0 in the best possible balance. For example, if you
image encryption—also known as ‘‘selective image
divide 10-byte plain text, the first five bytes are A0 and
encryption’’—is used to protect sensitive information or
the last five bytes are B0 . B0 is encrypted using the
sensitive information within an image. Partial image
advanced encryption standard (AES) encryption func-
encryption includes a method for selective encryption of
tion FK , along with the byte length n, tweak T , and
a specific spatial area31–36 and another method used to
round number of the message. The key uses a 128-bit
encrypt part of an image based on unique attributes such
AES key. The encrypted value is added to A0 , and then
as the frequency of the image.37,38
mod radix calculation is performed to make C0 . C0
becomes B1 , B0 becomes A1 , and a round ends. This
Format-preserving encryption process is repeated 10 times in total. Finally, when A10
and B10 are concatenated, the encryption process ends.
Format-preserving encryption is an encryption
approach in which the image data length is the same as
Contribution
that of the plain text and the form is that of the
cipher.39 Based on the prefix cipher, cycle-walking The proposed method has the following three
cipher, generalized Feistel cipher, rank-then-encryption advantages:
(RTE), and the tweak concept, practical approaches to
format-preserving encryption have been proposed.40,41  Specific spatial image partial encryption is
Standard format-preserving encryption methods that possible;
have been proposed include FF1 (FFX mode), FF2  The amount of image data does not increase as a
(VAES3), and FF3 (BPS).42–46 However, with the dis- result of padding;
covery of security vulnerabilities in FF2 and FF3, FF3-  Images can be used without exposing informa-
1, which is a modification of FF1 and FF3, has been tion requiring privacy.
established as the US National Institute of Standards
and Technology (NIST) standard.47,48 Standard Traditional image protection methods include pri-
format-preserving encryption requires three pieces of vacy masking, image full encryption, and image partial
4 International Journal of Distributed Sensor Networks

Figure 2. Encryption and decryption structure of FF1 and FF3-1.

encryption. Privacy masking incurs a problem in that it large image can adversely affect image recognition. In
cannot be decrypted. The image full encryption method contrast, the proposed method allows encryption of just
encrypts all the pixels so that the image cannot be the license plate and the person in the car. Therefore, it
recognized at all until after decryption, and once is possible to collect images while solving the problem
decrypted, information requiring privacy cannot be of invasion of citizens’ privacy. In addition, privacy
protected. Traditional image partial encryption masking is not decrypted, but the proposed method has
encrypts a large rectangular range that includes the area the advantage of allowing the decryption and use of an
requiring protection. This incurs a problem in that even encrypted area as needed (such as in the case of a crime
non-sensitive parts of the image are encrypted. In addi- committed in the street). The format-preserving encryp-
tion, the use of a block cipher causes problems because tion used for encryption can encrypt even short-length
it produces an image file that does not conform to the data without padding. Therefore, there is no increase in
structure determined by padding. The proposed method the amount of data due to padding.
performs encryption by targeting a pixel range that spe-
cifically covers an area that the user wishes to encrypt.
There are various methods for detecting objects to be Proposed method
encrypted in images, such as the use of deep learning,
Partial image encryption using format-preserving
OpenCV, or software.36,49–51 The data can be recog-
nized and used normally except for the encrypted area, encryption
such that images can still be used while protecting pri- This article proposes an image partial encryption
vate information. For example, if the traditional image method based on format-preserving encryption. It was
partial encryption was used to partially encrypt the implemented and tested using FF1 and FF3-1. The pro-
license plate of a parked vehicle, and the person inside, posed method, which can be used to select and encrypt
as captured by CCTV, the entire image or the vehicle’s a portion of an image while preserving the data size,
license plate and the person to be encrypted must be consists of three steps: (1) selection of the encryption
encrypted with a rectangle. Encrypting an unnecessarily area, (2) encryption of image pixel values using format-
Jang and Lee 5

preserving encryption, and (3) storage of encrypted Table 1. Definition of fields and functions used in the
image pixel values. First, to select the area to be encryption/decryption process of the proposed method.
encrypted in a video, the area that corresponds to pri-
Notation Description
vate information that can identify an individual, such as
a face, a part of a body, or a vehicle number, must be n Number of pixels in the selected area
identified. The X and Y coordinate values of all pixels in PixelnX X coordinate of the selected pixel data
the selected area are checked, and the pixel values are PixelnY Y coordinate of the selected pixel data
stored in the byte array. For example, a red–green–blue Pixelnvalue Value of the selected pixel data
(RGB)-format image stores three bytes of pixel values SelectPixel All Q concatenated data
per pixel, while a gray-format image stores one byte of m Length of SelectPixel divided by 24 bytes
PixelArraym Byte array divided into 24-byte units for encryption
pixel values per pixel. The proposed method can be and decoding
applied to both RGB- and gray-format images. The l Byte length of PixelArray
pixel values stored in the byte array are encrypted using Cm Result of encrypting/decoding PixelArraym
format-preserving encryption. When encrypting using EncPixelnX Value obtained by encrypting and decoding Pixelnvalue
format-preserving encryption, data can be divided by Tn Tweak values required for encryption
t Length of T
block size. In this article, the block size is set to 24 bytes K AES 128-bit secret key used for encryption
for FF1 and 32 bytes for FF3-1. Details of the block size ½Xs Representation of X as a byte string of s bytes
setting are described in section ‘‘Optimal block size.’’ AESK (X) Function for AES encryption of X using key K
The encryption process requires plain text, a key, and a PRF(X) Function to perform initialization vector operation
tweak. The plain text is the pixel value of the selected for CBC process
HTD(X) Function to convert hexadecimal data to decimal
area, and the key uses a 128-bit AES key. The tweak
data
uses the X and Y coordinates of all of the pixel data in DTHr (X) Function to convert decimal data to hexadecimal
the block to be encrypted. Format-preserving encryption data of length r
use a tweak in the encryption and decryption process,
unlike ordinary block ciphers. If the pixel values are AES: advanced encryption standard; CBC: cipher block chaining.
encrypted using the same tweak, one can visually infer
the original image because the same pixel value becomes
the same cipher text. In this case, the correlation between as the pixel values of the existing X and Y coordinates.
pixels may be vulnerable to known and selective plain As Figure 4 shows, only the area selected as the private
text attacks.52,53 To solve this problem, the tweak must information in the original image is encrypted. In addi-
be a constantly changing value. The proposed method tion, the private information can be encrypted and pro-
uses tweak as the value obtained by concatenating the X tected, and users can use encrypted images.
and Y coordinates of the pixel value to be encrypted.
The final step in the process is encrypting of all the pixel Application scenario
values in the selected area. The decryption process is the
The proposed method can be used in various applica-
reverse of the encryption process. Table 1 summarizes
tions such as CCTV camera systems and image drone
the fields and functions used in the encryption/decryp-
systems. Figure 5 is an activity diagram of a scenario in
tion process of the proposed method. Figure 3 shows the
which the proposed method is applied to a CCTV cam-
entire process of encrypting pixel values in an RGB-for-
era system.
mat image using the proposed method. Algorithm 1
As shown in the figure, a CCTV camera shoots the
shows the process of dividing the pixel value into prede-
video, which is then transmitted to the storage server.
termined block units before the encryption process of
The server encrypts the privacy information of the
the proposed method is applied. Algorithm 2 calculates
video received from the CCTV camera using the pro-
tweak T values used in fpeEncryption and fpeDecryption.
Algorithms 3 and 4 detail the fpeEncryption and posed method and stores the video in the server.
fpeDecryption processes of the proposed method, respec- Security officers and general users can view the videos
tively. Figure 4 presents an example of the encryption with privacy information encrypted by connecting to
process used in the proposed method. the server or over the Internet. Figure 6 shows a CCTV
As shown in Figure 3, we select the X and Y coordi- camera system incorporated with improved security
nates and the pixel values of the privacy information in and applicability provided by the proposed method.
the image. Then, a PixelArray is created by concatenat-
ing the selected pixel values in 24 bytes. PixelArray val- Experiment evaluation
ues are computed with the fpeEncryption and
fpeDecryption functions, respectively, according to
Experimental results
encryption/decryption. As a result, the pixel data The proposed method was applied to images (6.1.01,
EncPixel value encrypted in the generated C are stored 4.1.02, 5.2.08, house, 5.3.01, and 2.2.02) from the USC-
6 International Journal of Distributed Sensor Networks

Figure 3. Process sequence in the proposed method.

SIPI database of various formats and sizes. Figure 7 Comparison with traditional image-security methods
shows the original image, while Figures 8–10 show the The proposed method was compared with the tradi-
results of full image encryption using AES, FF1, and tional image-security methods to assess the possibility
FF3-1, respectively. Figures 11 and 12 show the results of decryption, the degree to which file sizes increase,
of encrypting only the information requiring privacy in and the possibility of recognizing encrypted images.
the image using the proposed method. The following problems arise when the traditional
image-security approach is applied to an image pro-
cessing system. Privacy masking prevents decoding
Increasing data size in videos because the original data are lost by modifying pixel
A video is data in which multiple frames are collected. values. Full image encryption using a block encryption
When a captured video is encrypted in real time, the method, such as AES, does not permit encryption of a
image data size may increase due to padding. Table 2 part of an image; the entire image must be encrypted.
summarizes the increase in the data size over time when Therefore, the image cannot be used until decoding is
all the frames of a captured video are encrypted in real performed, and private information is exposed when
time. The data size increases as a result of the addition the decoding is performed. In addition, because of the
of 16 bytes padding data to every frame under various characteristics of the block code, the size of the image
frame conditions. The units are Mb. is invariably increased by padding. Figure 13 shows the
For CCTV, which involves the continuous recording results of encryption using privacy masking, image full
of video, the size of the video may increase over time encryption, and the proposed method.
such that storage space is wasted. As the number of Figure 13(b) shows that private information can be
CCTV cameras capturing video increases, the wastage inferred when mosaic masking is used and is not
of server storage space worsens. decrypted. Figure 13(b) shows mosaic image using the
Jang and Lee 7

Figure 4. Partial image encryption process using format-preserving encryption.

Algorithm 1. PixelArray schedule Algorithm 2. Tweak schedule 1

1: procedure PixelArray (Pixelvalue , n) 1: procedure T (PixelnX , PixelnY )

2: m = dn=8e 2: m = dn=8e
3: for i in 1to(m  1) do 3: for i in 1tom do
4: q = (i  1)38 4: q = (i  1)38
5: PixelArryai = SelectPixel½q + 1    q + 8 5: Ti = PixelqX+ 1 k PixelqY + 1 k    k PixelqX+ 8 k PixelqY + 8
6: end for 6: end for
7: PixelArraym = SelectPixel½8(m  1) + 1    n 7: Return T = T1 k T2 k    k Tm
8: Return PixelArray = PixelArray1 k PixelArray2 k    k PixelArraym 8: end procedure
9: end procedure

the proposed method. The image remains recognizable

pixelate of Adobe Photoshop. Figure 13(c) shows the while the private information in the image is protected,
image encrypted for all pixel data using AES-128 CBC and the data size of the image does not increase. If
mode. The encrypted image is unrecognizable, and the video is encrypted by applying the proposed method in
data size is increased by padding. Figure 13(d) and (e) a CCTV camera system, both the monitoring control
shows the result of encrypting the image by applying room and the users connected to the Internet can use
8 International Journal of Distributed Sensor Networks

Algorithm 3. Encrypt function

1: procedure fpeEncrypt (K, T, PixelArray)

2: u = bl=2c; v = l  u
3: A = PixelArray½1    u; B = PixelArray½u + 1    l
4: b = ddv3LOG(radix)ee=8
5: d = 4db=4e + 4
6: P = ½11 k ½21 k ½11 k ½radix3 k ½101 k
½u mod(256)1 k ½n4 k ½t4
7: for i in 0 to 9 do
8: Q = T k ½0(tn1)mod16 k ½i1 k ½HTD(B)b
9: R = PRF(P k Q)
10: S be the first d bytes of the following
string of dd=16e block: Figure 5. Activity diagram of application scenario in CCTV
11: R k AESK (R  ½116 ) k AESK (R  ½216 ) camera system with the proposed method.
k    k AESK (R  ½d=16  116 )
12: y = HTD(S)
13: if i is even then
14: r=u
the images while private information remains protected,
15: else without decrypting the data. In general, no decryption
16: r=v is required, so no key exchange is required, and because
17: end if no padding is required for encryption, storage space is
18: c = ðHTD(A) + yÞmod(16)r not wasted. Table 3 compares the results of the pro-
19: O = DTHr (c)
20: A=B
posed method and traditional image-security methods
21: B=O in three respects: whether image decryption is permit-
22: end for ted, whether and how file size is increased, and whether
23: Return A k B encrypted images can be recognized.
24: end procedure The proposed method solves the problems associ-
ated with privacy masking (i.e. that encrypted images
cannot be decrypted) and encryption using block
ciphers (i.e. increased file size and wasted storage
space). The proposed method improves security by
Algorithm 4. Decrypt function encrypting only private information in the image and
improves the availability of images to users by provid-
1: procedure fpeDecrypt (K, T, PixelArray) ing image information other than private information.
2: u = bl=2c; v = l  u
3: A = PixelArray½1    u; B = PixelArray½u + 1    l
4: b = ddv3LOG(radix)e=8e
5: d = 4db=4e + 4
Performance
6: P = ½11 k ½21 k ½11 k ½radix3 k ½101 k Speed
½u mod(256)1 k ½n4 k ½t4
7: for i in 9 to 0 do We verified the performance of the proposed scheme by
8: Q = T k ½0(tn1)mod16 k ½i1 k ½HTD(A)b measuring the encryption/decryption rates for Figures
9: R = PRF(P k Q) 8–12. The experimental results are listed in Table 4.
10: S be the first d bytes of the following
string of dd=16e block:
11: R k AESK (R  ½116 ) k AESK (R  ½216 ) Optimal block size
k    k AESK (R ½dd=16e  116 )
12: y = HTD(S) To improve the encryption/decryption speed of the pro-
13: if i is even then posed scheme, we determined the optimal block size
14: r=u when encrypting with the FF1 and FF3-1 algorithms.
15: else FF1 and FF3-1, as used in the proposed method, do
16: r=v not have a fixed block length because they must be able
17: end if
18: c = ðHTD(B) + y Þmod(16)r to encrypt various lengths. As the time required to
19: O = DTHr (c) encrypt an image varies depending on the length of the
20: B=A block, the length of a block suitable for encryption/
21: A=O decryption should be set. Figure 14 is a graph showing
22: end for the encryption times for FF1 and FF3-1 according to
23: Return A k B
24: end procedure block length using Figure 7(c) and (d). As the length of
the block increases, the encryption speed of FF1 also
Jang and Lee 9

Figure 6. CCTV camera system equipped with proposed method.

Figure 7. Original image: (a) 6.1.01, (b) 4.1.02, (c) 5.2.08, (d) house, (e) 5.3.01, and (f) 2.2.02.

Figure 8. Result of full encryption using AES: (a) 6.1.01, (b) 4.1.02, (c) 5.2.08, (d) house, (e) 5.3.01, and (f) 2.2.02.

Figure 9. Result of full encryption using FF1: (a) 6.1.01, (b) 4.1.02, (c) 5.2.08, (d) house, (e) 5.3.01, and (f) 2.2.02.
10 International Journal of Distributed Sensor Networks

Figure 10. Result of full encryption using FF3-1: (a) 6.1.01, (b) 4.1.02, (c) 5.2.08, (d) house, (e) 5.3.01, and (f) 2.2.02.

Figure 11. Result obtained with proposed method using FF1: (a) 6.1.01, (b) 4.1.02, (c) 5.2.08, (d) house, (e) 5.3.01, and (f) 2.2.02.

Figure 12. Result obtained with proposed method using FF3-1: (a) 6.1.01, (b) 4.1.02, (c) 5.2.08, (d) house, (e) 5.3.01, and (f) 2.2.02.

Table 2. Data size increase due to padding in full encryption.

Table 3. Comparison of privacy masking, image encryption,
1 hour 1 day 1 week 1 month and the proposed method.
24 fps 0.0219 0.5273 3.6914 110.7421 Privacy AES Our method
25 fps 0.0228 0.5493 3.8452 115.3564 masking (FF1, FF3-1)
30 fps 0.0274 0.6591 4.6142 138.4277
50 fps 0.0457 1.0986 7.6904 230.7128 Decryption 3 O O
60 fps 0.0549 1.3183 9.2285 276.8554 Awareness O 3 O
100 fps 0.0915 2.1972 15.3808 461.425 of image
120 fps 0.1098 2.6367 18.4570 553.710 File size No change Increase No change

AES: advanced encryption standard.

increases, but it is slightly slower at 25 bytes. FF1 was encrypts the plain text into 24-byte units and FF3-1
the fastest from a length of 25 + (32 3 a) bytes (e.g. encrypts it into 32 bytes.
25, 57, 89). FF3-1 gets faster as the block length
increases. Because of the characteristics of the algo-
rithm, the block length of FF3-1 was set to a maximum Security evaluation
of 32 bytes and tested. As a result, it can be seen that
32 bytes is the most efficient block length. Therefore, in To evaluate the security of the proposed method, histo-
the encryption process of the proposed method, FF1 gram, correlation coefficient, number of pixels change
Jang and Lee 11

Figure 13. Image encryption results: (a) original image (4,147,254 bytes), (b) privacy masking (4,147,254 bytes), (c) AES (4,147,270
bytes), (d) proposed method (FF1, 4,147,254 bytes), and (e) proposed method (FF3-1, 4,147,254 bytes).

Table 4. AES and proposed rate of encryption and decryption.

Encryption (ms) Decryption (ms)

Figure 8 8(a) 3.243099 3.216401

8(b) 2.449099 2.410699
8(c) 2.697301 2.6238
8(d) 6.769 5.9916
8(e) 3.248401 3.354301
8(f) 9.5483 9.546301
Figure 9 9(a) 172.5461 39.0605
9(b) 527.3073 14.93308
9(c) 689.3037 4.533195
9(d) 2076.729 196.1894
9(e) 2759.026 1554.101
9(f) 8283.313 846.7711
Figure 10 10(a) 196.2087 43.6244
10(b) 589.5579 15.54559
10(c) 783.7156 5.231196
10(d) 2346.49 317.8314
10(e) 3131.515 1642.126
10(f) 9394.513 949.9847
Figure 11 11(a) 39.21519 176.96918 Figure 14. Encryption speed by block length for Figure 7(c)
11(b) 14.821 533.02506 and (d).
11(c) 4.549602 706.27681
11(d) 195.3813 2125.5758
11(e) 1571.194 2828.251
11(f) 844.1251 8472.7691 FF1, and FF3-1 all produce balanced histogram
Figure 12 12(a) 44.39931 198.293 graphs.
12(b) 15.78791 594.2996
12(c) 5.242199 795.5995
12(d) 325.2316 2380.341
12(e) 1663.161 3145.791 Correlation coefficient analysis
12(f) 958.1904 9462.259 Correlation coefficient analysis indicates the indepen-
dence of two images through correlation calculation
between pixels. If the correlation value is linear, the
rate (NPCR), unified average change intensity (UACI), independence of the two images is not good. The more
and information entropy analyses, and NIST SP 800-22 nonlinear the correlation values, the more independent
tests were performed. the two images. In other words, the higher the correla-
tion between the two images, the closer the absolute
value of the weight is to 1; conversely, the higher the
Histogram analysis independence, the closer the absolute value of the
weight is to 0. Formula (1) is used to calculate the cor-
A histogram analysis graphically illustrates the weights relation value, where x and y are the mean pixel values
of the pixel values in an image. The more uniform the of the two images and N is the total number of pixels
histogram distribution value, the better the security.
Figure 15 shows the histogram results for Figures 7–12. cov(x, y)
The histogram distributions of the original images are Cxy = pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð1Þ
D(x)D(y)
all unbalanced. However, for full encryption, AES,
12 International Journal of Distributed Sensor Networks

Figure 15. Results of histogram analysis.

!
1 X
N NPCR and UACI analysis
cov(x, y) = (xi  E(x))(yi  E(y)) ð2Þ
N i=1 The resistance to a chosen plain text attack (differential
attack) was evaluated using NPCR and UACI for an
1X N
image encrypted using the proposed method and block
E(x) = xi ð3Þ
N i=1 cipher encryption. NPCR is the pixel variation rate
between the plain text and the encrypted image, and
1X N UACI is the average variation intensity of plain text
D(x) = (xi  E(x))2 ð4Þ and the encrypted image. This analysis method can
N i=1
reveal whether the change in the original image affects
encryption. If the original image is greatly changed in
Table 5 shows the correlation coefficient weight val-
comparison with the encrypted image, it can be said to
ues for RGB and gray images with respect to the origi-
have good resistance to a chosen plain text attack (dif-
nal image for AES and the proposed method. These
ferential attack). NPCR and UACI are calculated using
results show that the proposed method yields a correla-
formulas (5)–(7)54,55
tion coefficient weight value similar to those obtained
by image encryption with AES. P
i, j D(i, j)
NPCR = 3 100% ð5Þ
w3h
Jang and Lee 13

Table 5. Results of correlation coefficient analysis.

(a) (b) (c) (d) (e) (f)

Figure 8 0.002395 –0.001916 0.000593 –0.00109 0.000823 0.000004

0.00027 0.000195 –0.000716
0.001921 0.003237 0.000339
Figure 9 0.006383 0.002356 0.001665 0.001308 0.00128 –0.000138
0.000646 0.001087 0.00259
–0.004169 0.000969 –0.001693
Figure 10 –0.00001 0.000321 0.0019 –0.001468 –0.000271 –0.00011
0.00008 –0.000582 0.000779
0.004872 –0.00176 –0.001176
Figure 11 0.739148 0.913697 0.987642 0.841272 0.173474 0.658663
0.846253 0.855386 0.142743
0.835704 0.895915 –0.063458
Figure 12 0.73817 0.915152 0.987152 0.840426 0.173112 0.658453
0.848129 0.855665 0.142069
0.838938 0.896612 –0.062387

Table 6. Results of NPCR analysis.

(a) (b) (c) (d) (e) (f)

Figure 8 0.996201 0.996302 0.995956 0.995983 0.996097 0.996067

Figure 9 0.996445 0.996353 0.996178 0.995975 0.996034 0.99608
Figure 10 0.996246 0.995987 0.996136 0.996045 0.996038 0.996078
Figure 11 0.226669 0.026886 0.006577 0.094377 0.527941 0.101272
Figure 12 0.226501 0.026855 0.006573 0.094407 0.527862 0.101279

Table 7. Results of UACI analysis.

(a) (b) (c) (d) (e) (f)

Figure 8 0.320871 0.403926 0.275398 0.308853 0.324934 0.301467

Figure 9 0.320275 0.403527 0.275884 0.30892 0.324853 0.301292
Figure 10 0.321293 0.403677 0.275876 0.308864 0.32536 0.301332
Figure 11 0.06659 0.009809 0.001703 0.028268 0.193876 0.033252
Figure 12 0.06606 0.009716 0.001745 0.028272 0.193986 0.033242

!
1 X jC1 (i, j)  C2 (i, j)j through image encryption using block cipher (AES). In
UACI = 3 100% ð6Þ other words, the proposed method is as safe as image
w 3 h i, j F
encryption using a block cipher (AES).

1 C1 (i, j) 6¼ C2 (i:j)
D(i, j) ð7Þ
0 C1 (i, j) = C2 (i:j)
Information entropy analysis
In formula (6), C1 denotes a plain image, C2 denotes Information entropy analysis is used to measure the
an encrypted image, and w and h are the horizontal and complexity of the encrypted data. The encrypted data
vertical sizes of the image, respectively. F is a value that must be complex enough to prevent information on the
represents the maximum support pixel value compatible original data from being obtained. The optimal entropy
with the image format of the cipher text, and the image value of the RGB and gray images used is 8. Table 8
of the 8-bit format is 255. Tables 6 and 7 present the lists the entropy values for the original images in RGB
NPCR and UACI values, respectively, obtained for and gray formats and those achieved by encrypting the
images in RGB and gray formats with Figures 8–12. images using the proposed method and full encryption.
These results show that the proposed method yields The results in Table 8 show that the proposed
NPCR and UACI values similar to those obtained method and image encryption using a block cipher
14 International Journal of Distributed Sensor Networks

Table 8. Results of information entropy analysis.

(a) (b) (c) (d) (e) (f)

Figure 7 7.2044467 6.2944978 7.201008 7.4857872 7.5237369 6.5785567

Figure 8 7.9971571 7.9989656 7.9991992 7.9997215 7.999797 7.9999211
Figure 9 7.9978015 7.9990454 7.9993489 7.9997852 7.9998212 7.9999405
Figure 10 7.9972393 7.9990452 7.9993289 7.9997645 7.9998074 7.9999412
Figure 11 7.3347614 6.4000599 7.218075 7.5697277 7.858972 6.8655011
Figure 12 7.3287123 6.3991863 7.2180681 7.5699342 7.8589302 6.8651993

Table 9. Results of NIST SP 800-22 tests.

Figure 11(e) Figure 11(f) Figure 12(e) Figure 12(f) Pass/Fail

Frequency test 0.5641175 0.3561378 0.5163866 0.7413325 All pass

Frequency test within a block 0.7400311 0.6041753 0.5178028 0.5233779 All pass
Runs test 0.5730168 0.103916 0.7509988 0.6392064 All pass
Test for the longest run of ones in a block 0.4921633 0.6371338 0.6157612 0.0386303 All pass
Binary matrix rank test 0.2662403 0.7795677 0.8560301 0.1314836 All pass
Discrete Fourier transform test 0.1750349 0.3813249 0.9341601 0.7652656 All pass
Non-overlapping template matching test 0.9996462 1.0000137 1 1 All pass
Overlapping template matching test 0.4720836 0.4837326 0.257433 0.369052 All pass
Maurer’s ‘‘Universal statistical’’ test 0.8827247 0.1445012 0.4011447 0.8781729 All pass
Linear complexity test 0.1087466 0.3251907 0.3641394 0.2987047 All pass
Serial test 0.7523306 0.131774 0.0125111 0.1323363 All pass
Approximate entropy test 0.8934906 0.1823651 0.0125726 0.4899735 All pass
Cumulative sums test 0.4193308 0.4878375 0.4555566 0.8067146 All pass
Random excursions test 0.4600447 0.0400129 0.0354376 0.2340035 All pass
Random excursions variant test 0.1367363 0.0518082 0.1243495 0.2208814 All pass

(AES) yield values close to the optimal entropy. The distribution of the histogram and correlation
Therefore, both the proposed method and the image coefficient of the proposed method is good, and the
encryption using a block cipher (AES) are sufficiently results of NPCR and UACI are very close to those
complex. obtained with encryption using AES. In addition, the
value of the information entropy is as close to the opti-
mal value as that of the AES-encoded image. Finally,
NIST SP 800-22 tests all the NIST SP 800-22 tests were passed. Therefore,
NIST SP 800-22 tests were used to test the cipher ran- the proposed method can be said to be a safe means of
using both FF1 and FF3-1.
domness.56 The goal of those tests is to analyze the ran-
domness of the encrypted image data. The analysis
consisted of 15 tests. For the experimental results Conclusion
obtained in this study, tests were performed on the data
for the encrypted region in the 1024 3 1024 size image In this article, we proposed a method for partially
of Figures 11 and 12, and all the tests were passed to encrypting private information, such as faces and body
verify the randomness. The test results are listed in parts, in images using the format-preserving encryption
Table 9. standards FF1 and FF3-1. Traditional image protec-
tion technology has problems such as increasing data
by padding and storage space wastage over time.
Security analysis comparison Furthermore, as the entire image is encrypted, the
image cannot be recognized before decryption and,
To evaluate the security level of images encrypted with once decrypted, information requiring privacy is
AES and with the proposed method, histogram, corre- exposed. Traditional partial image encryption incurs a
lation, NPCR, UACI, information entropy analysis, problem in that unnecessary portions are encrypted by
and NIST SP 800-22 were performed as described encrypting a rectangular area that covers the informa-
above. tion requiring privacy. The proposed method solves this
Jang and Lee 15

problem. The proposed method has no padding, so Proceedings of the 5th ACM symposium on computer
there is no increase in the data size, and the information human interaction for management of information technol-
requiring privacy can be specifically set and encrypted. ogy, Cambridge, MA, 4–5 December 2011, p.4. New
In this study, we measured the encryption and decryp- York: ACM.
6. Cho JH and Ahn CW. Auto detection system of personal
tion speed of the proposed method and determined the
information based on images and document analysis. J
most suitable block unit for encryption to improve the
Inst Internet Broadcast Commun 2015; 15(5): 183–192.
encryption and decryption speed of the image part. The 7. Bang J, Kang S and Kim MS. The study of factors to
security of the proposed method was evaluated through affect on users’ self-disclosure in social networking ser-
analysis of histograms, correlation coefficients, NPCR, vices. J Korea Acad Ind Cooperation Soc 2016; 17(8):
UACI, information entropy, and NIST SP 800-22. We 69–76.
verified through experiments that the proposed method 8. Kumari S. A research paper on cryptography encryption
is as secure as encryption using AES. The proposed and compression techniques. Int J Eng Comput Sci 2017;
method can be used in various environments that use 6(4): 20915–20919.
images, such as Internet of things (IoT), drones, and 9. De Choudhury M, Sundaram H, Lin YR, et al. Connect-
ing content to community in social media via image con-
CCTV, and can be useful in responding to image leak-
tent, user tags and user communication. In: 2009 IEEE
age threats, such as privacy invasion and confidential international conference on multimedia and expo, New
information leakage. York, 28 June–3 July 2009, pp.1238–1241. New York:
IEEE.
Declaration of conflicting interests 10. Yeung CMA, Kagal L, Gibbins N, et al. Providing access
control to online photo albums based on tags and linked
The author(s) declared no potential conflicts of interest with data. In: AAAI spring symposium on social semantic web:
respect to the research, authorship, and/or publication of this where Web 2.0 meets Web 3.0, Palo Alto, CA, 23–25
article. March 2009, pp.9–14. Menlo Park, CA: AAAI.
11. Pesce JP, Casas DL, Rauber G, et al. Privacy attacks in
Funding social media using photo tagging networks: a case study
with Facebook. In: Proceedings of the 1st workshop on
The author(s) disclosed receipt of the following financial sup- privacy and security in online social media, Lyon, 17 April
port for the research, authorship, and/or publication of this 2012, p.4. New York: ACM.
article: This work was supported by the National Research 12. Christin D, López PS, Reinhardt A, et al. Share with
Foundation of Korea (NRF) grant funded by the Korean strangers: privacy bubbles as user-centered privacy con-
government (MSIT; No. 2018R1A4A1025632), the Basic trol for mobile content sharing applications. Inform Secur
Science Research Program through the NRF that is funded Tech Rep 2013; 17(3): 105–116.
by the Ministry of Education (No. NRF- 13. Mannan M and Van Oorschot PC. Privacy-enhanced
2018R1D1A1B07047656), and the Soonchunhyang sharing of personal content on the web. In: Proceedings
University. of the 17th international conference on World Wide Web,
Beijing, China, 21–25 April 2008, pp.487–496. New York:
ORCID iD ACM.
14. Vyas N, Squicciarini AC, Chang CC, et al. Towards auto-
Wonyoung Jang https://orcid.org/0000-0002-2777-9830
matic privacy management in Web 2.0 with semantic
analysis on annotations. In: 2009 5th international confer-
References ence on collaborative computing: networking, applications
and worksharing, Washington, DC, 11–14 November
1. Yu J, Zhang B, Kuang Z, et al. iPrivacy: image privacy
2009, pp.1–10. New York: IEEE.
protection by identifying sensitive objects via deep multi-
15. Squicciarini AC, Shehab M and Paci F. Collective pri-
task learning. IEEE Trans Inf Foren Sec 2016; 12(5):
vacy management in social networks. In: Proceedings of
1005–1016.
the 18th international conference on World Wide Web,
2. Wang C, Zhang B, Ren K, et al. Privacy-assured out-
Madrid, 20–24 April 2009, pp.521–530. New York:
sourcing of image reconstruction service in cloud. IEEE
ACM.
Trans Emerg Top Comput 2013; 1(1): 166–177.
16. Klemperer P, Liang Y, Mazurek M, et al. Tag, you can
3. Nov O, Naaman M and Ye C. Motivational, structural
see it! Using tags for access control in photo sharing. In:
and tenure factors that impact online community photo Proceedings of the SIGCHI conference on human factors
sharing. In: Third international AAAI conference on in computing systems, Austin, TX, 5–10 May 2012,
Weblogs and social media, San Jose, CA, 17–20 May pp.377–386. New York: ACM.
2009. Menlo Park, CA: AAAI. 17. Ravichandran R, Benisch M, Kelley PG, et al. Capturing
4. Squicciarini AC, Xu H and Zhang X. CoPE: Enabling social networking privacy preferences. In: International
collaborative privacy management in online social net- symposium on privacy enhancing technologies symposium,
works. J Am Soc Inf Sci Technol 2011; 62(3): 521–534. Seattle, WA, 5–7 August 2009, pp.1–18. Berlin: Springer.
5. Wang N, Xu H and Grossklags J. Third-party apps on 18. Bonneau J, Anderson J and Church L. Privacy suites:
Facebook: privacy and the illusion of control. In: shared privacy for social networks. In: Proceedings of the
16 International Journal of Distributed Sensor Networks

5th symposium on usable privacy and security (SOUPS multimedia applications (SIGMAP), Barcelona, 28–31
2009), vol. 9, Mountain View, CA, 15–17 July 2009, July 2007, pp.324–329. New York: ACM.
pp.1–2. New York: ACM. 35. Parameshachari BD, Karappa R, Soyjaudah KS, et al.
19. Bonneau J, Anderson J and Danezis G. Prying data out Partial image encryption algorithm using pixel position
of a social network. In: 2009 international conference on manipulation technique: the smart copyback system. In:
advances in social network analysis and mining, Athens, 2014 4th international conference on artificial intelligence
20–22 July 2009, pp.249–254. New York: IEEE. with applications in engineering and technology, Kota
20. Fang L and LeFevre K. Privacy wizards for social net- Kinabalu, Malaysia, 3–5 December 2014, pp.177–181.
working sites. In: Proceedings of the 19th international New York: IEEE.
conference on World Wide Web, Raleigh, NC, 26–30 April 36. Panduranga HT and Naveenkumar SK. Selective image
2010, pp.351–360. New York: ACM. encryption for medical and satellite images. Int J Eng
21. Park SC. Design and implementation of personal infor- Technol 2013; 5(1): 115–121.
mation identification and masking system based on image 37. Abdmouleh MK, Khalfallah A and Bouhlel MS. A novel
recognition. J Inst Internet Broadcast Commun 2017; selective encryption scheme for medical images transmis-
17(5): 1–8. sion based-on JPEG compression algorithm. Procedia
22. Kaur M and Kumar V. A comprehensive review on Comput Sci 2017; 112: 369–376.
image encryption techniques. Arch Comput Method Eng 38. Parameshachari BD, Panduranga HT and Naveenkumar
2018; 27: 15–43. SK. Partial encryption of medical images by dual DNA
23. Younes MAB. A survey of the most current image addition using DNA encoding. In: 2017 international con-
encryption and decryption techniques. Int J Adv Res ference on recent innovations in signal processing and
Comput Sci 2019; 10(1): 9. embedded systems (RISE), Bhopal, India, 27–29 October
24. OVSENÍ KĹ, Turán J, Huszanı́k T, et al. Image encryp- 2017, pp.310–314. New York: IEEE.
tion algorithm with plaintext related chaining. Comput 39. Brightwell M and Smith H. Using datatype-preserving
Inf 2019; 38(3): 647–678. encryption to enhance data warehouse security. In: 20th
25. Brindha K, Sharma R and Saini S. Use of symmetric national information systems security conference proceed-
algorithm for image encryption. Int J Innov Res Comput ings (NISSC), Baltimore, MD, 7–10 October 1997,
Commun Eng 2014; 2(5): 4401–4407. pp.141–149. Gaithersburg, MD: NIST.
26. Gupta K, Gupta R, Agrawal R, et al. An ethical 40. Black J and Rogaway P. Ciphers with arbitrary finite
approach of block based image encryption using chaotic domains. In: Cryptographers’ track at the RSA confer-
map. Int J Secur Appl 2015; 9(9): 105–122. ence, San Jose, CA, 13–17 February 2002, pp.114–130.
27. Zhang Y. A fast image encryption algorithm based on Berlin: Springer.
convolution operation. IETE J Res 2017; 65(1): 4–18. 41. Bellare M, Ristenpart T, Rogaway P, et al. Format-pre-
28. Li Y, Wang C and Chen H. A hyper-chaos-based image serving encryption. In: International workshop on selected
encryption algorithm using pixel-level permutation and areas in cryptography, Calgary, AB, Canada, 13–14
bit-level permutation. Opt Lasers Eng 2017; 90: 238–246. August 2009, pp.295–312. Berlin: Springer.
29. Oravec J, Turan J, Ovsenı́k L, et al. Asymmetric image 42. Bellare M, Rogaway P and Spies T. The FFX mode of
encryption approach with plaintext-related diffusion. operation for format-preserving encryption. Gaithersburg,
Radioengineering 2018; 27(1): 281–288. MD: NIST, 2010.
30. Lagmiri SN, Elalami N and Elalami J. Color and gray 43. Vance J. VAES3 scheme for FFX: an addendum to the
images encryption algorithm using chaotic systems of dif- FFX mode of operation for format preserving encryption.
ferent dimensions. Int J Comput Sci Network Secur 2018; Gaithersburg, MD: NIST, 2011.
18(1): 79–86. 44. Brier E, Peyrin T and Stern J. BPS: a format-preserving
31. He J, Liu B, Kong D, et al. PUPPIES: transformation- encryption proposal. Ɋɟɠɢɦ ɞɨɫɬɭɩɭ ɞɨ ɪɟɫɭɪɫɭ, 2010.
supported personalized privacy preserving partial image http://csrc.nist.gov/groups/ST/toolkit/BCM/documents/pro
posedmodes/bps/bps-spec.pdf
sharing. In: 2016 46th annual IEEE/IFIP international
45. Dworkin MJ and Perlner RA. Analysis of VAES3 (FF2),
conference on dependable systems and networks (DSN),
2015, https://eprint.iacr.org/2015/306.pdf
Toulouse, 28 June–1 July 2016, pp.359–370. New York:
46. Dworkin M. Recommendation for block cipher modes of
IEEE.
operation: methods for format-preserving encryption. NIST
32. Parameshachari BD, Kiran Rashmi P, Supriya MC, et al.
Special Publication, 800-38G, 2016, https://nvlpubs.nist.-
Controlled partial image encryption based on LSIC and
gov/nistpubs/SpecialPublications/NIST.SP.800-38G.pdf
chaotic map. In: Proceedings of the 3rd international con-
47. Durak FB and Vaudenay S. Breaking the FF3 format-
ference on cryptography, security and privacy, Kuala Lum-
preserving encryption standard over small domains. In:
pur, Malaysia, 19–21 January 2019, pp.60–63. New York:
Annual international cryptology conference, Santa Bar-
ACM.
bara, CA, 20–24 August 2017, pp.679–707. Cham:
33. Steinebach M, Liu H, Stein R, et al. Hybrid image
Springer.
encryption. Electron Imaging 2018; 2018(7): 1–6.
48. Dworkin M. Recommendation for block cipher modes of
34. Wong A and Bishop W. Backwards compatible, multi-
operation: methods for format-preserving encryption.
level regions-of-interest (ROI) image encryption architec-
NIST Special Publication, 800-38G, Revision 1, 2019,
ture with biometric authentication. In: Proceedings of the
https://nvlpubs.nist.gov/nistpubs/SpecialPublications/
second international conference on signal processing and
NIST.SP.800-38Gr1-draft.pdf
Jang and Lee 17

49. He K, Gkioxari G, Dollár P, et al. Mask R-CNN. In: Proceedings of the Korean information science society con-
Proceedings of the IEEE international conference on com- ference, 2012, pp.274–276. Seoul, Korea: Korean Insti-
puter vision, Venice, 22–29 October 2017, pp.2961–2969. tute of Information Scientists and Engineers.
New York: IEEE. 53. Pareek NK, Patidar V and Sud KK. Diffusion-substitu-
50. Su H, Wei S, Yan M, et al. Object detection and instance tion based gray image encryption scheme. Digital Signal
segmentation in remote sensing imagery based on precise Process 2013; 23(3): 894–901.
mask R-CNN. In: IGARSS 2019: 2019 IEEE interna- 54. Chen G, Mao Y and Chui CK. A symmetric image
tional geoscience and remote sensing symposium, Yoko- encryption scheme based on 3D chaotic cat maps. Chaos
hama, Japan, 28 July–2 August 2019, pp.1454–1457. New Soliton Fract 2004; 21(3): 749–761.
York: IEEE. 55. Mao Y, Chen G and Lian S. A novel fast image encryp-
51. Zhou Y, Panetta K, Cherukuri R, et al. Selective object tion scheme based on 3D chaotic baker maps. Int J Bifurc
encryption for privacy protection. In: Mobile multimedia/ Chaos 2004; 14(10): 3613–3624.
image processing, security, and applications 2009, vol. 56. Bassham LE III, Rukhin AL, Soto J, et al. A statistical
7351, Orlando, FL, 13–17 April 2009, p.73510F. Belling- test suite for random and pseudorandom number generators
ham, WA: International Society for Optics and Photonics. for cryptographic applications. Sp 800-22 rev. 1a, 2010,
52. Chung JY and Hong YS. Parallel image encryption https://csrc.nist.gov/publications/detail/sp/800-22/rev-1a/
schemes for security of ultra high resolution images. In: final