Securing Image Transmission Using in- Compression Encryption Technique

Shaimaa A. El-said, Khalid F. A. Hussein, & Mohamed M. Fouad
International Journal of Computer Science and Security, (IJCSS), Volume (4): Issue (5) 466
Securing Image Transmission Using In- Compression
Encryption Technique
Shaimaa A. El-said Eng.sahmed@windowslive.com
Faculty of Engineering / Electronics
and Communication Department
Zagazig University
Zagazig,44519, Egypt.
Khalid F. A. Hussein khalid_elgabaly@yahoo.com
Electronics research institute
Microwaves Department
Researches National Institute
Dokki, Egypt
Mohamed M. Fouad fouadzu@hotmail.com
Faculty of Engineering / Electronics
and Communication Department
Zagazig University
Zagazig,44519, Egypt.
Abstract
Multimedia is one of the most popular data shared in the Web, and the protection
of it via encryption techniques is of vast interest. In this paper, a secure and
computationally feasible Algorithm called Optimized Multiple Huffman Tables
(OMHT) technique is proposed. OMHT depends on using statistical-model-based
compression method to generate different tables from the same data type of
images or videos to be encrypted leading to increase compression efficiency and
security of the used tables. A systematic study on how to strategically integrate
different atomic operations to build a multimedia encryption system is presented.
The resulting system can provide superior performance over other techniques by
both its generic encryption and its simple adaptation to multimedia in terms of a
joint consideration of security, and bitrate overhead. The effectiveness and
robustness of this scheme is verified by measuring its security strength and
comparing its computational cost against other techniques. The proposed
technique guarantees security, and fastness without noticeable increase in
encoded image size.
Keywords: Image encryption and compression, optimized multiple Huffman tables, OMHT performance
analysis, computational cost analysis
1. INTRODUCTION
With the rapid development of multimedia and network technologies, the security of multimedia
becomes more and more important, since multimedia data are transmitted over open networks
more and more frequently. Typically, reliable security is necessary to content protection of digital
images and videos. Encryption schemes for multimedia data need to be specifically designed to

protect multimedia content and fulfill the security requirements for a particular multimedia
application. For example, real-time encryption of an entire video stream using classical ciphers
requires heavy computation due to the large amounts of data involved, but many multimedia
applications require security on a much lower level, this can be achieved using selective
encryption that leaves some perceptual information after encryption.
As an important way of designing a secure video encryption schemes, secret Multiple Huffman
Tables (MHT) have been suggested in some designs. The major advantage by using this kind of
joint compression-encryption approach is that high compression ratio and high encryption degree
can be achieved in one single step, which simplifies the system design and makes it flexible for
some advanced multimedia processing [1] in addition to the reduction of time required to perform
compression followed by encryption. After re-studies the security of multimedia encryption
scheme based on secret Huffman tables, the present cryptanalysis shows presence of drawbacks
in MHT technique.
To overcome the drawbacks of MHT technique, a new scheme for more general and efficient
secure multimedia transmission, OMHT, is proposed. OMHT depends on using statistical-model-
based compression method to generate different tables from a training set has the same data
type as images or videos to be encrypted leading to increase compression efficiency and security
of the used tables. Using known fixed tables in MHT technique generated by mutation (a method
introduced in [1]) for compressing and encrypting images causes degradation in both
compression ratio and security. We focus our research attention to enhancing multiple Huffman
tables coding techniques. It is a challenging problem to verify joint consideration of security,
bitrate overhead, and friendliness to delegate processing. Performance analysis of the newly
proposed scheme OMHT shows that it can provide superior performance over both generic
encryption and MHT in the security and compression.
This paper is organized as follows: Section 2 shows an overview of multimedia encryption
techniques. A new proposed scheme, Optimized Multiple Huffman tables coding technique
(OMHT) is described in section 3 with a detailed description for proposed adaptive quantization
technique. Section 4 presents a performance analysis of the proposed scheme OMHT technique.
The computational cost of the proposed technique is analyzed in section 5. Conclusion is
given in section 6.
2. OVERVIEW of MULTIMEDIA ENCRYPTION TECHNIQUES
When dealing with still images, the security is often achieved by using the naïve (traditional)
approach to completely encrypt the entire image, traditional encryption, with a standard cipher [2]
(DES, AES, IDEA, etc.). As shown in Fig. (1), assuming that the plaintext and the ciphertext are
denoted by P and C, respectively, the encryption procedure in a cipher can be described as C = E
Ke (P), where Ke is the encryption key and E(·) is the encryption function. Similarly, the
decryption procedure is P =DKd (C), where Kd is the decryption key and D(·) is the decryption
function When Ke = Kd, the cipher is called a private-key cipher or a symmetric cipher For
private-key ciphers, the encryption-decryption key must be transmitted from the sender to the
receiver via a separate secret channel. When Ke ≠ Kd, the cipher is called a public-key cipher or
an asymmetric cipher. For public-key ciphers, the encryption key Ke is published, and the
decryption key Kd is kept private, for which no additional secret channel is needed for key
transfer. Ciphering the complete compressed file may result in excessive computational burden
and power consumption at the decoder and perhaps even the server/ encoder.
However, there are number of applications for which the naive based encryption and decryption
represents a major bottleneck in communication and processing. Some recent works explored a
new way of securing the content, named, partial encryption or selective encryption, soft
encryption, perceptual encryption, by applying encryption to a subset of a bitstream. The main
goal of selective encryption is to reduce the amount of data to encrypt while achieving a required
level of security [3].

encryption decryption
kdKe
ceiphertextplaintext Recovered
plaintext
Figure 1: Traditional Encryption Techniques
Transform/
prediction
quantization Entropy
coding
Multiplex
/packetize
2 3 41 5
FIGURE 2: Candidate Domains Used to Apply Encryption to Multimedia.
According to Fig. 2, there are two straight forward places to apply generic encryption to
multimedia. The first possibility is to encrypt multimedia samples before any compression, stages
1 and 2, Qiao et al. [4] and Uehara and Safavi-Naini [5] are examples of pre-compression
selective encryption. The main problem with this approach is that the encryption often significantly
changes the statistical characteristics of the original multimedia source, resulting in much reduced
compressibility. Cheng and Li, 2000. The wavelet-based compression algorithm SPIHT [6] is an
example of post-compression encryption scheme, stage 4 and 5. Wu et al proposed encryption
scheme based on encoding with multiple Huffman tables (MHT) used alternately in a secret order
[1]; is an example of in-compression selective encryption stages 3, and 4. The encryption with
reasonably high level of security and unaffected compression can be achieved simultaneously,
requiring almost negligible additional overhead. One of the major advantages by using this kind of
joint encryption-compression approach is that encryption and compression can be achieved in
one single step, which simplifies the system design an makes it flexible for some advanced
multimedia processing such as scalability and rate shaping.
2.1. Multiple Huffman Tables (MHT) Technique
The MHT algorithms [1][7]-[9], aiming to increase the model space while maintaining the
computational efficiency, keep the structure of the Huffman tree but enlarge the model space
through tree mutation. The procedure of the basic MHT algorithm is described as follows:
Step1: Train four original Huffman trees from different sets of training data. e.g. Huffman table of
the JPEG DC coefficients.
Step2: Based on the original trees, perform tree mutation, to create the whole Huffman tree
space.
Step3: Randomly select m different tables from the space, and number them from 0 to m-1.
Step4: Generate a random vector P = {P0,P1,·· ,Pm-1} each p is an Integer ranging from 0 to m-1.
Step5: For the i
th
encountered symbol, use table Pi (modn) to encode it.
MHT coding [1] makes use of standard coding tables. It is included in the final bit-stream for every
image to be compressed. This approach presents disadvantages:
1. Visual degradation: very high-visual degradation can be achieved.

2. Cryptographic security: Gillman and Rivest [10] showed that decoding a Huffman coded
bitstream without any knowledge about the Huffman coding tables would be very difficult.
However, the basic MHT is vulnerable to known and chosen plaintext attacks as pointed out
in [11].
3. It writes all codes of the corresponding tables in the final bitstream even if only some of they
were used to encode the associated events of the particular input image.
4. It does not make use of any statistic about the distribution of the events of the image.
To improve the security several kinds of enhanced MHT schemes have been proposed:
• By inserting random bit in the encrypted bit stream or integrating with a stream cipher [8].
• Recently another scheme via random rotation in partitioned bit streams has been
reported [9].
3. OPTIMIZED MULTIPLE HUFFMAN TABLES (OMHT)
OMHT compression-encryption technique is a modification to the MHT scheme; it generates
different Huffman tables for each type of images instead of using fixed Huffman tables for all
images as in MHT technique. The main advantage of using OMHT technique over other lossy
compression technique is that it produces a much smaller compressed file than any compression
method, while still meeting the advantage of encryption. Remove small, invisible parts, of the
picture is based on an accurate understanding of how the human brain and eyes work together to
form a complex visual system. As a result of these subtle reductions, a significant reduction in the
resultant file size for the image sequences is achievable with little or no adverse effect in their
visual quality. As shown in Fig. 3 OMHT process takes two parallel paths A, and B, so it takes no
additional time to add encryption to the compressed bitstream as both traditional and selective
encryption techniques.
3.1. The Procedure of Compressing the Original Image
As shown in Fig. 3 (path A), The input NxM image is first converted into single vector by
concatenating successive rows beside each other to form a long row that contains all the image
pixels using matrix to vector converter. This vector is exposed to DCT to transform the image
from spatial domain into frequency domain in which energy of the image information is
concentrated in a few number of coefficients. The output of the DCT process is a vector that has
the same length of the image (number of pixels in the image), but with many values approximated
to zeros. After applying the DCT the output coefficients are arranged in a descending order
according to its energy content. The energy content of the coefficients is summed from the
beginning of the vector and toward the end till a specific energy percentage (EP) of the image
energy is reached. Those coefficients that carry EP energy percent are chosen to be transmitted
and the rest coefficients are neglected since they carry only very small energy that will not affect
the visual quality of the recovered image. This EP value depends on image characteristics and it
can be varied to achieve the desired compression ratio and the signal to noise ratio according the
application: As we decrease the EP value, a higher compression ratio is obtained with slightly
lower signal to noise ratio. Now the number of the transmitted coefficients (Tc) becomes very
small. The reduced coefficients vector returned back to the spatial domain using IDCT to be
processed by an efficient quantizer.
The proposed Least Probable Coefficients Approximation (LPCA) quantizer as shown in Fig. 4(a)
and its flowchart in Fig. 4(b) reduces the output values of the IDCT by calculating their occurrence
probabilities. The IDCT coefficients are arranged in a descending order according to their
probabilities in a vector. The desirable quantization levels are taken as the most probable
coefficients from the beginning of the arranged vector; if the required CR and SNR are achieved
by transmitting only four quantization levels, those quantization levels are the first four coefficients
in the arranged vector. The probability of the last QL is called neglecting probability (NP).

FIGURE 3: Optimized Multiple Huffman Table (OMHT) Coding System
All coefficients with probability less than Np are approximated to the nearest quantization level in
value. This technique is irreversible; this means that the dequantized values can’t be turned back
to their original values leading to quantization losses. The quantization procedures are
summarized as follows:
• IDCT coefficients are arranged in a descending order according to their probabilities in a
vector.
• The desirable n quantization levels are taken as the n most probable coefficients from the
beginning of the arranged vector.
• The probability of the last QL is called neglecting probability (NP).
• All coefficients with probability less than Np are approximated to the nearest quantization level
in value. The proposed quantization reduces number of transmitted values but not the number
of transmitted coefficients.
After the transmitted values are reduced by quantization, each quantized level is assigned a
codeword using Huffman encoder that enables representing an image in a more efficient way with
smallest memory for storage or transmission. Huffman coding is used to code the quantized
values statistically according to their probability of occurrences. Short code words are assigned to
highly probable values and long code words to less probable values. The average number Lavg of
bits required to represent a symbol is defined as,
∑=
=
L
k
kkavg rPrIL
1
)()(
Where, rk is the discrete random variable for k=1,2,…L with associated probabilities P(rk). The
number of bits used to represent each value of rk is I(rk). The number of bits required to
represent an image is calculated by number of symbols multiplied by Lavg [9].
OMHT decoding
Dequantization
Transmission over channel
OR Storage on a disk
Encoding using
OMHT
DCTCoefficient
recovery
IDCTReshape
image
random M
subsets selector
Generate Table
from each subset
Datasets
database
dataset
selector
Secret key
genration
B
A
M
Image to
vector
QuantizationCoefficients
selector
IDCTDCT
Recovered
image
Original
image

(a)
(b)
FIGURE 4: (a) Block Diagram of LPCA Quantizer (b) LPCA flowchart
3.2. Procedure of Preparing and Using the OMHT Tables
Following is the procedure of preparing and using the optimized multiple Huffman tables and how
it is used to both encode and encrypt images as shown in Fig. 3 (path B).
Step 1: images training set are divided into L datasets. Each dataset's images have the same
properties.
Step 2: each dataset contains N images.
Step3: The input image compared to datasets to select the dataset that has the same properties.
Step4: randomly choose M subsets each subset contains K images from the dataset.
Concatenate all images of each subset and calculates the pixels probabilities. Then draw
Huffman tree and find the Huffman table contains the different pixels' values and their
associated variable codewords. Now we have M different tables to be used.
Step5: Tables are saved at each decoder, and the order by which the tables are generated and
used is kept secret.
Step6: Number the generated M tables from 0 to M-1.
Step7: Generate a random vector P (the secret order) its length equal to the length of image
under consideration. Each element value in P ranges from 0 to M-1.
Step8: For the ith encountered symbol (coefficient to be encoded), use table P i(mod n) to
encode it.
Coefficients
ordering
TP
determinator
n
Coefficients
rounding off
Quantizer
Quantized
coefficients
IDCT
coefficients
If i>n
yes
i=i+1
end
Probabilities descending
ordering
Choose n desirable
number
of QL
TP=Prob.(U(I))
start
Read IDCT coefficients
vector (U)
Qu(i)=U(i)
no
i=0

4. PERFORMANCE ANALYSIS of OMHT
For performance evaluation, the following experiments measure the compression performance
and encryption strength of OMHT using test images that contains gray and colored images. The
compression performance of OMHT is analyzed by calculating the compression ratio (CR),
number of bits per symbol (BPP), the peak signal to noise ratio (PSNR), and the mean square
error (MSE). A comparison between the proposed scheme based on generating tables based on
statistical modeling of large dataset for each types of images, and a compression using fixed
predetermined encoding tables, JPEG standard, on which the MHT technique based on done to
show the effectiveness of the proposed scheme in compression. The encryption strength of the
OMHT is tested and compared with other encryption techniques.
The achieved compression ratio can be calculated from the following equation:
compressed
original
CR =
Where the original is the size of the original image and the compressed is the size of the Huffman
encoder output compressed bitstream. Calculate the bit per pixel (BPP) is defined as:
P
B
BPP =
Where P is the total number of pixels in an image and B is the total number of transmitted bits for
this image. As a measure of reconstructed image quality, the peak signal-to-noise ratio (PSNR) in
dB is used, this is defined as follows:
MSE
PSNR
n
dB
12
log20 10
−
=
Both mean square error (MSE) and the signal to noise ratio (SNR) for an nXn image are
calculated from the following equations:
( )
2
1 1
],[],[
2
1
∑∑= =
−
=
n
i
n
j
jiji
n
MSE βα
( )
( )
∑∑
∑∑
= =
−
= =
×= n
i
n
j
jiji
n
i
n
j
ji
SNR
1 1
2],[],[
1 1
],[
10
2
log10
βα
α
Where, α[i,j] and β[i,j] denote the original and decoded levels of the pixel [i,j] in the image,
respectively. A larger PSNR value means that the encoded image preserves the original image
quality better.
Experiment 1 uses lossy OMHT to encrypt and compress the Lena image. It gives the ability to
control the compression ratio and peak signal to noise ratio by either change number of QL while
the amount of Tc is constant, or changing the amount of Tc while number of QL is constant.
As shown in Table 1, and Fig. 5, and 6, while the number of quantization levels is constant at
q=128 and the amount of transmitted DCT’s coefficients changes from Tc=98.5% of the image
energy to Tc= 99.9%. As Tc increases, the CR decreases providing an increase in PSNR.

(a) Original Image (b) Decoded for TC=99.9% (c) Decoded for TC= 99.6% (d) Decoded for TC=99.4%
(e) Decoded forTC=99.2% (f) Decoded for TC= 99% (g) Decoded for TC= 98.8% (h) Decoded for TC=98.5%
FIGURE 5: The Effect of Reducing Number of Transmitted Coefficients on Image Visual
Degradation
Tc=99.9%Tc=99.6%Tc=99.4%Tc=99.2%Tc=99%Tc=98.8%Tc=98.5%
CR 2.8179 7.0518 10.860 15.252 20.612 26.936 38.973
BPP 2.8390 1.1345 0.7366 0.5245 0.3881 0.2970 0.2074
PSNR34.032 28.744 26.7053 25.929 24.794 24.065 23.125
MSE 20.848 94.222 116.673 179.287 200.80 290.092 360.09
SNR 26.947 21.669 19.6207 18.8446 17.719 16.9809 16.040
TABLE 1: Compression Performance of Applying OMHT with Different Number of Transmitted
Coefficients
98.5 99 99.5 100
0
5
10
15
20
25
30
35
40
Tc%
CR
OMHT
98.5 99 99.5 100
22
24
26
28
30
32
34
36
Tc%
PSNR
OMHT
(a) Variation of CR with Varying Tc (b) Variation of PSNR with Varying Tc
FIGURE 6: The Effect of Reducing OMHT Number of Tc on CR, and PSNR
OMHT provides the ability to maintain Tc constant and varies the number of QL. As shown in
Table 2, and Fig. 7, and 8, while the amount of transmitted coefficients is constant at Tc=99.5%
and the number of quantization levels changes from using four quantization levels to using 256
quantization levels. As the number of quantization levels increases, the compression ratio
decreases providing an increase in peak signal to noise ratio.

(a) Original Image (b) Decoded for q=256 (c) Decoded for q=128 (d) Decoded for q=64
(e) Decoded for q=32 (f) Decoded for q=16 (g) Decoded for q=8 (h) Decoded for q=4
FIGURE 7: The Effect of Reducing Number of Quantization Levels on Image Visual Degradation
Q=4 Q=8 Q=16 Q=32 Q=64Q=128Q=256
CR 24.0818.43715.9913.25 11.4510.49 10.14
BPP 0.3320.43390.5000.604 0.6990.762 0.789
PSNR22.0923.29524.4226.14 27.5927.75 27.92
MSE 382.2305.51222.2119.67102 98.12 80.94
SNR 15 16.21 17.3319.06 20.5 20.66 20.84
TABLE 2: Compression Performance of Applying OMHT with Different QL on Lena Image
0 50 100 150 200 250 300
10
15
20
25
QL
CR
OMHT
0 50 100 150 200 250 300
22
23
24
25
26
27
28
QL
PSNR
OMHT
(a) Variation of CR with Varying QL (b) Variation of PSNR with Varying QL
FIGURE 8: The Effect of Reducing OMHT Number of Quantization Levels on CR, And PSNR
Experiment 2 compares the compression performance of lossy OMHT with that of lossy JPEG
technique to prove that the proposed technique adds security without affecting the compression
ratio or the PSNR. Table 3 provides a comparison of CR between OMHT, and JPEG at Different
BPP on Lena Image, while Table 4 provides a comparison of PSNR between OMHT, and JPEG
at Different BPP on Lena Image. From Tables 3 and 4 it is obvious that using lossy OMHT
technique provides higher PSNR and storage space and transmission bandwidth required than
JPEG especially at low bitrates.

BPP OMHT JPEG
0.2
39 39.01
0.18
44.4 43
0.16
49.4 49
0.14
57.4 58.02
0.12
65.6 65.3
0.1
73.9 73
TABLE 3: Comparison of CR between OMHT, and JPEG at Different BPP on Lena Image
BPP OMHT JPEG
0.2
22.6 21.14
0.18
22.2 20
0.16
21.9 19.4
0.14
21.6 18
0.12
21.3 16.7
0.1
21 15
TABLE 4: Comparison of PSNR between OMHT, and JPEG at Different BPP on Lena Image
0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18 0.19 0.2
35
40
45
50
55
60
65
70
75
BPP
CR
OMHT
JPEG
0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18 0.19 0.2
15
16
17
18
19
20
21
22
23
BPP
PSNR
OMHT
JPEG
FIGURE 9: CR of both OMHT technique FIGURE 10: PSNR of both OMHT technique and
and JPEG for Lena image JPEG for Lena image
Fig. 9 shows that both OMHT technique and JPEG technique have nearly the same compression
levels at the same number of bits per pixel for Lena image. Fig. 10 shows that PSNR of OMHT
technique is higher and more stable at low bitrate than that of JPEG for Lena image.
Experiment 3 measures the encryption strength performance of the proposed OMHT technique,
colored football image in RGB (288x352x3) shown in Fig. 11(a) with its histogram shown in Fig.
11(b) is compressed and encrypted using the OMHT uses multiple Huffman tables, generated
from a large set of training images that have the same type of the test image used in a secret
order (secret key). Fig. 11(c) and 11(d) shows the test image and its histogram after decoding it
with another technique as JPEG. While Fig. 11(e) and 11(f) shows the test image and its

histogram after decoding it with OMHT technique and the same encoding tables but without
knowing the secret order (secret key). It is obvious that OMHT provides high perceptual security
-50 0 50 100 150 200 250 300
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
(a) Original Image (b) Histogram of Original Image
-50 0 50 100 150 200 250 300
0
0.5
1
1.5
2
2.5
x 10
4
(c) Decoded image with JPEG(PSNR=5dB) (d) Histogram of decoded Image by JPEG
-50 0 50 100 150 200 250 300
0
500
1000
1500
2000
2500
3000
(e) Decoded with OMHT wrong key (PSNR=5.7dB) (f) Histogram of decoded Image by OMHT
FIGURE 11: The Effect of Decoding Football Image without the Secret Order
Fig.12 shows the perceptual performance comparison between OMHT and other different
encryption techniques used to encrypt Lena image. Fig.12(a) shows the original Lena image,
Fig.12(b) shows the decoded image that was encrypted by OMHT, Fig.12(c) shows the decoded
image that was encrypted by building a three level pyramid and encrypting the lowest resolution
plus the first residual (HP Mode 30% encryption), Fig.12 (d) shows the decoded image that was
encrypted by encrypting only the DC coefficients with the first AC coefficient of each block (SS
Mode 30%), Fig.12 (e) shows the decoded image that was encrypted by scrambling the DC
coefﬁcients and one bitplane or three bitplanes (MM Mode 30%), Fig.12(f) shows the decoded
image that was encrypted by encrypting the most significant bits of all coefficients (SA Mode
30%), Figs.2.band3.b clearly show that there can be still information left in the unencrypted parts
of the data after selective encryption has been applied, Fig.12 (g) shows the decoded image that
was Encrypted by Run-length, Fig.12 (h) Encrypted by sign bit encryption, Fig.12 (i) shows the
decoded image that was Encrypted by band permutation(10 bands), Fig.12(j) shows the decoded
image that was encrypted by bitplane permutation(n=6), Fig.12(k) shows the decoded image that
was encrypted by bitplane permutation (n=7), and Fig.12(l) shows the decoded image that was
encrypted by MHT. It is obvious that the PSNR of the decoded image that is encrypted by OMHT

is smaller than it is in all other techniques. So, the perceptual security strength of the OMHT
technique is higher than other techniques.
(a) (b) (c)
(d) (e) (f)
(g) (h) (i)
(j) (k) (l)
FIGURE 12: The Effect of Decoding Lena Image Encrypted with Different Techniques by JPEG:
(a) Original Image, (b) Encrypted by OMHT(PSNR=4.8 dB), (c) Encrypted by HP Mode
(PSNR=14.7 dB), (d) Encrypted by SS Mode(PSNR=14.2 dB), (e) Encrypted by MM Mode
(PSNR=6.2 dB), (f) Encrypted by SA Mode(PSNR=6.4 dB), (g) Encrypted by Run-length
(PSNR=6.5 dB), (h) Encrypted by sign bit encryption (PSNR=6.1 dB), (i) Encrypted by band
permutation(10 bands) (PSNR=7.23 dB) (j) Encrypted by bitplane permutation (n=6) (PSNR=13.8
dB), (k) Encrypted by bitplane permutation (n=7) (PSNR=9.18 dB), (l) Encrypted by MHT
(PSNR=6.4 dB),

5. COMPUTATIONAL COST ANALYSIS
The evaluation of the computational speed of ciphers usually consists of the analysis of the key-
setup cost, the encryption cost and the decryption cost [16]. The encryption and the decryption
costs are usually similar, and they are more important than the key-setup cost because one
single key-setup can often be followed by thousands of encryption/decryption operations. In the
following, we analyze these costs of our OMHT encryption scheme, and compare them with those
of MHT and modern ciphers.
a) Key-Setup cost: The key-setup process includes all the computation and memory allocation
operations prior to actual encryption of the first bit in the plaintext. The computational cost of
OMHT key-setup is dominated by the construction of optimized multiple Huffman tables,
generation of the secret order by which those tables are used, and comparing the test image with
datasets. OMHT takes about 10 operation per table generation, single operation for secret key
generation, and L operation for comparison. The total number of operations equal 10XMXL+1+L,
where L, M is number of datasets and number of subsets respectively. For L=4, M=20, the net
Key-Setup cost =805 operations. For MHT technique it takes 20 operations per table entry, the
total cost would be 20xtxm, where t and m are the table size and the number of selected tables,
respectively. For the example of JPEG dc coefficient encryption as shown in the previous
subsection, the key-setup cost would be around 2000 operations ( t=13 and m=8 ).Compared
with the ciphers listed in Table 6,the key-setup cost of OMHT encryption is much smaller than
MHT and other ciphers.
b) Encryption/Decryption cost: The net computational cost of the OMHT is the same as the
basic MHT-encryption scheme [1] is less than one CPU operation per encrypted bit as explained
below. When a symbol is to be encoded with a normal Huffman coder, the shift amount is added
to the base address of the table to obtain the address of the desired Huffman code. This process
is illustrated in Fig.13 (a). In the basic MHT system, we store the base addresses of the tables in
a cyclic queue according to the order that they are used. When a symbol is to be
encoded/encrypted, the base address is first loaded from the memory, and then the shift-amount
is added to it. Afterwards, the index to the cyclic queue of base addresses should be increased by
one. Then, the index should be compared with the end of the queue in order to decide whether it
should be reset to the beginning of the queue. Therefore, the computational difference between
our cipher/encoder and a normal Huffman coder is one memory-load, one addition and one
comparison operation for each symbol encoded. The encoding process of the proposed
cipher/encoder is shown in Fig.13 (b). Since each symbol in the original data usually corresponds
to more than 3 bits in the Huffman bitstream, then encryption cost of our algorithm is less than
one CPU operation per encrypted bit, which is around 20 times smaller than the well-known AES
as listed in Table 6.
Recently, a new cryptographic cipher named COS [18] with a very fast speed is gaining
popularity. It is around 4–5 times faster than AES. Compared to COS, the encryption cost of
OMHT is still several times smaller.

TABLE 6: Computational Costs of AES Finalists on a Pentium-MMX Machine. The Figures in
This Table are Translated from [17] by Assuming Two CPU Instructions are Executed in Every
Clock Cycle in a Pentium-MMX CPU
(a) (b)
FIGURE 13: (a) Normal Huffman Coder Adds the Shift Amount to the Base address of the Table
to Obtain the Address of the Desired Huffman Code. (b) OMHT Loads the Base Addresses of
Huffman Tables from a Cyclic Queue, and the Index to the Queue is Increased by One After
Coding of Each Symbol.
6. CONCLUSIONS
The experiments’ results reveal that the proposed OMHT technique achieves better compression
and security performance than that of MHT, and JPEG Image Compression Standard especially
at low bitrate. The OMHT scheme provides
• High security: resistance against various types of attacks, including the ciphertext-only
attack and the known/chosen plaintext attack[19].
• Low encryption cost: the encryption cost not exceed very small portion of the total
computation cost of compression
• No harm to the compression ratio: The increase of the final bit stream size due to
encryption is not higher than 0.5% of the original coded bitstream.
• Joint compression-encryption OMHT technique achieves both high security and compression
performance in one single step, which simplifies the system design and reduces time required
to perform compression followed by encryption.
• Since images have different statistics, using the same fixed JPEG standard predefined
coding tables as suggested in MHT technique will not be effective in encoding all image and
video types.

• The OMHT method obtains better performance in terms of storage space use and more
stable peak signal to noise ratio than that of JPEG in encoding an image with small and great
gray-level variations among adjacent pixels.
• Receivers haven't the secret order cannot decode the encoded images successfully.
• Further, the proposed new compression-encryption technique could be applied on any source
data, not only images, which uses Huffman coding to achieve better compression ratio.
Therefore, the proposed technique will be suitable for compression of text, image, and video
files.
7. REFERANCES
1. C.-P. Wu and C.-C. J. K. Kuo. “Design of integrated multimedia compression and encryption
systems”. IEEE Transactions in Multimedia, vol. 7, no. 5, pp. 828–839, 2005.
2. W. Stallings. "Cryptography and Network Security Principles and Practices", Upper Saddle
River, NJ: Prentice Hall, 2003.
3. M. Van Droogenbroeck and R. Benedett. “Techniques for a selective encryption of
uncompressed and compressed images”. In Proceedings of Advanced Concepts for
Intelligent Vision Systems (ACIVS '02), pp. 90–97, Ghent, Belgium, September 2002.
4. L. Qiao, K. Nahrstedt, and M.-C. Tam.“Is MPEG encryption by using random list instead of
zigzag order secure?”. in Proceedings of the IEEE International Symposium on Consumer
Electronics (ISCE '97), pp. 226–229, Singapore, December 1997.
5. T. Uehara and R. Safavi-Naini.“Chosen DCT coefficients attack on MPEG encryption
scheme”. in Proceedings of IEEE Pacific Rim Conference on Multimedia, pp. 316–319,
Sydney, Australia, December 2000.
6. H. Cheng and X. Li. “Partial encryption of compressed images and videos”. IEEE
Transactions on Signal Processing, vol. 48, no. 8, pp. 2439–2451, 2000.
7. C.-P. Wu and C.-C. Kuo. “Efficient multimedia encryption via entropy codec design”. Proc.
SPIE, vol. 4314, Jan. 2001.
8. D. Xie and C. J. Kuo. “Enhanced Multiple Huffman Table (MHT) Encryption Scheme Using
Key Hoping”. In Proceedings of IEEE International Symposium on Circuits and Systems,
pp.568–571, May2004.
9. D. Xie and C. J. Kuo. “Multimedia Data Encryption via Random Rotation in Partitioned Bit
Stream”. In Proceedings of IEEE International Symposium on Circuits and Systems, pp.568–
571, May2004.
10. D. W. Gillman and R. L. Rivest. “On breaking a Huffman code”. IEEE Transactions on
Information Theory, vol. 42, no. 3, pp. 972–976, 1996.
11. J. Zhou, Z. Liang, Y. Chen, and O. C. Au. “Security analysis of multimedia encryption
schemes based on multiple Huffman table”. IEEE Signal Processing Letters, vol. 14, no. 3,
pp. 201–204, 2007.
12. W. Pennebaker and J. Mitchell. ”JPEG Still Image Data Compression Standard”, Van
Nostrand Reinhold, New York, 1993.
13. http://www.jpeg.org (JPEG resources) [accessed at 4/8/2010]

14. http://www.jpeg.org/public/jfif.pdf (JPEG file interchange format) [accessed at 8/8/2010]
15. (independent JPEG group) ftp.uu.net:/graphics/jpeg [accessed at 8/8/2010]
16. C.-P. Wu and C.-C.J. Kuo. “Efficient multimedia encryption via entropy codec design”. In
Proc. SPIE Int. Symp. Electronic Imaging 2001, vol. 4314, Jan. 2001, p.128.
17. J. Nechvatal et al. “Report on the Development of the Advanced Encryption Standard”.
National Institute of Standards and Technology, U.S. Dept. Commerce, Tech. Rep., Oct.
2000.
18. E. Filiol and C. Fontain. “A new ultra fast stream cipher design: COS ciphers”. In Proc. 8
th
IMA Conf. Cryptography and Coding, Dec. 2001.
19. Shaimaa A. El-said, Khalid F. A. Hussein, and Mohamed M. Fouad. “Securing Multimedia
Transmission Using Multiple Huffman Tables Technique”. Electrical and Computer Systems
Engineering Conference (ECSE’10), Egypt, 2010.

Securing Image Transmission Using in- Compression Encryption Technique

More Related Content

What's hot (16)

Similar to Securing Image Transmission Using in- Compression Encryption Technique (20)

Recently uploaded (20)

Securing Image Transmission Using in- Compression Encryption Technique