Page 1 of 13

Transactions on Engineering and Computing Sciences - Vol. 11, No. 3

Publication Date: June 25, 2023

DOI:10.14738/tecs.113.14919.

Ching-Yu, Y., Tsung-Hsiang, H., & Pei-Yun, C. (2023). A New ECG Steganography Based on Enhanced Coefficient Adjustment

Technique and Least Significant Bit Substitution. Transactions on Engineering and Computing Sciences, 11(3).122-134.

Services for Science and Education – United Kingdom

A New ECG Steganography Based on Enhanced Coefficient

Adjustment Technique and Least Significant Bit Substitution

Ching-Yu, Yang

Dept. Computer Science and Information Engineering,

National Penghu University of Science and Technology, Taiwan

Tsung-Hsiang, Hsu

Dept. Computer Science and Information Engineering,

National Penghu University of Science and Technology, Taiwan

Pei-Yun, Chen

Dept. Computer Science and Information Engineering,

National Penghu University of Science and Technology, Taiwan

ABSTRACT

A new data hiding technique was presented to conceal the privacy of patient’s

diagnosis and personal sensitive information in an electrocardiogram (ECG) signal.

Instead of hiding one data bit in a host block at a time by traditional coefficient

adjustment technique (CAT), we embed two data bits in the same size of the block

via the proposed enhanced CAT (ECAT) method. The resultant payload provided by

ECAT is nearly twice as large as that provided by the CAT scheme. In addition, if

ECAT equipped with the least significant bit (LSB) substitution, then a large number

of bits can be hidden in the ECG hosts while the resulting perceived quality is good.

Simulations revealed that both the average payload and SNR and the proposed

method are superior to those of the CAT-based schemes and existed ECG

steganographic methods. The application of our method can be found in portable

(or wearable) measurement devices.

Keywords: Data hiding, ECG steganography, CAT, ECAT, LSB.

INTRODUCTION

Due to steady growth of very high-speed backbone network service, the improving chip

manufacturing technology and ubiquitous use of powerful applications in mobile devices, plus

the maturity of artificial intelligence algorithms and 5G communication technology, people can

easily roam from the Internet and share their resources conveniently. However, data can be

eavesdropped or tampered with during transmission. Data hiding [1-3] provides a simple and

economical ways to protect (or secure) sensitive or important information. Generally speaking,

data hiding can be classified as steganography and digital watermarking [4-6]. A major goal of

steganography is to achieve hiding storage as large as possible while the resultant perceived

quality is good. Since a good perceived quality of marked media attracts no attention from the

third parties. Typical applications of steganography can be found in covert channel

communications between two parties and the hidden of private (or important) data in

multimedia. As for a major purpose of digital watermarking is to purse robustness performance

Page 2 of 13

123

Ching-Yu, Y., Tsung-Hsiang, H., & Pei-Yun, C. (2023). A New ECG Steganography Based on Enhanced Coefficient Adjustment Technique and Least

Significant Bit Substitution. Transactions on Engineering and Computing Sciences, 11(3).122-134.

URL: http://dx.doi.org/10.14738/tecs.113.14919.

while maintaining acceptable perceptual quality. Namely, the watermarking schemes intend to

protect copyrights, hoping to avoid or prevent intentionally (or unintentionally) manipulating

of watermarked multimedia. Recently, researchers pay attention to hide patient’s diagnosis

and personal data in biometric measure signals such as electrocardiogram (ECG) and

Electroencephalography [7-13]. Since the objective of the proposed method is to conceal secret

message in an ECG host and improve the performance of the variants of CAT-based ECG

steganography [7,10, 12, 13], only related (lossy) methods are reviewed here.

Yang and Wang [7] used coefficient alignment and developed a high-quality and a high-capacity

ECG steganography. Simulation revealed that the average SNR/ payload were 54.00 dB/ 7,500

bits and 42.47/ 14,783 for the high-quality and high-capacity versions, respectively. To

improve previous work, Yang and Wang [10] suggested an absolute-value-decision policy to

hide secrete bits in the ECG hosts. According to the decision policy, at most two data bits can

be hidden in a host block via the adjustment of the coefficients in the block. In addition, to

obtain a high-perceived quality, a 1-bit version of the method was presented to achieve the

goal. An apparent merit of the method is to use no extra information during the bit

embedding/extraction process. Simulations showed that the average SNR (in dB)/ payload (in

bits) of the method was 47.54/ 19,614 and 57.80/ 9,982 for the 2-bit and 1-bit version of the

method, respectively. Based on two-dimensional (2D) approach, Yang et al. [12] proposed an

ECG steganography. Before conducting bit embedment, they converted input ECG data from

one-dimensional (1D) arrangement into series of 2D block with the size of 3  3. According to

the coefficient adjustment approach, six data bits can be hidden in a host block. Simulations

indicated that the average SNR of the method is superior to that of existed techniques, whereas

the payload of the method is competitive to that of the compared methods. An apparent

disadvantage of the method was the waste of residual blocks (without being used for bit

hidden) after the conversion of ECG from 1D to 2D.

Based on the integer wavelet transform (IWT), Yang and Wang [13] presented a data hiding to

hide personal message in an ECG host via the offset of IWT coefficients. To embed data bits in

the coefficients of a host block with the size of 1  n, they used the coefficient adjusting rule,

that is, |

∑i=0

n−2

sji

n−1

− sj(n−1)

| ≤ τ, where τ is a control integer, n is the size of the host block, and sji

is the ith coefficient of the jth block. Namely, a data bit 1 or 0 can be hidden in a host block of

size 2 if −τ ≤ (sj0 − sj1) < 0 or 0 ≤ (sj0 − sj1) < τ is satisfied. Subsequently, input the next

bit 1 or 0 can be embedded in the block of size 3, if either −τ ≤ (

sj0−sj1

2

− sj2) < 0 or 0 ≤

(

sj0−sj1

2

− sj2) < τ is satisfied. Theoretically, data bits (of length n − 1) can be concealed in a host

block of size n. It is obvious that a large value of n provides a large hiding space, but degrades

the perceived quality (with the lower SNR) by the method. The procedure of coefficient

adjustment has to be conducted if rule-violation occurs during bit embedding. In addition, if a

block (or sub-block) carries no data bit, it is referred to as a skipped block. No extra information

is required for the recoding of the position of skipped blocks, because the skipped blocks can

be easily detected at the receiver according to the above coefficient adjusting rule.

Page 3 of 13

124

Transactions on Engineering and Computing Sciences (TECS) Vol 11, Issue 3, June - 2023

Services for Science and Education – United Kingdom

PROPOSED METHOD

To improve the hiding capacity of traditional CAT [13], an enhanced ECAT was presented here

to achieve the goal. Instead of hiding two data bits in a host bundle with the size of 1  3, the

proposed ECAT method embeds four bits in a host block with the same size. Theoretically, the

coding efficiency (4 / 3) of the ECAT and is twice as large as that (2 / 3) of the CAT. Specifically,

at most four data bits can be embedded in a bundle according to the following two decision

rules, namely, R1 and R2. They are defined as follows:

R1: |xi − xi+1

| < 2τ (1)

and

R2: |

xi−xi+1

2

− xi+2| < 2τ (2)

where  is a control parameter, xi, xi+1, and xi+2 are the three adjacent coefficients in a host

bundle. Without loss of generality, let α, β, γ, and δ be represented as the two bits “00,” “01,”

“10,” and “11,” which come from input bit-stream. The proposed ECAT method consists of two

phases. In phase I, α, β, γ, and δ can be “virtually” hidden into a host bundle of size 2 if R1 is

satisfied. In phase II, two data bits can be “carried” by the bundle of size 3 if R2 is satisfied. If

none of the above rules is satisfies, then the bundle carries no data bit and is referred to as a

skipped block. To further increasing hiding space, the ECAT equipped with the LSB substitution

is presented here. That is, instead of skipping the bundle (or sub-bundle), we embed data bits

in the skipped bundles by the LSB. Namely, in Phase I, we embed four data bits in the

coefficients of xi and xi+1 via the LSB as 2τ ≤ |xi− xi+1| is true. However, to avoid the altering the

values of the two coefficients (marked by Phase I), only two bits would be hidden into the

coefficient xi+2 via the LSB (in Phase II) if the condition of 2τ ≤ |

(xi+xi+1)

2

− xi+2| is satisfied. The

main steps of bit embedding and bit extraction for the proposed ECAT with the LSB method is

summarized in the following sections.

Bit Embedding

The main procedure of bit embedding is described in the following algorithm.

Algorithm 1: Embedding a secret message in an ECG host.

Input: Host ECG H, the size of a bundle n, control integer τ, and secret message W.

Output: Marked ECG Ĥ .

Method:

Step 1. Input a bundle, say Hj = {xi

}i=0

n−1

from the jth bundle of Η. If the end of input is

encountered, then go to Step 14.

Step 2. Read in two data bits  ∈ {00, 01, 10, 11} from W and compute the offset η = xi − xi+1.

Page 4 of 13

125

Ching-Yu, Y., Tsung-Hsiang, H., & Pei-Yun, C. (2023). A New ECG Steganography Based on Enhanced Coefficient Adjustment Technique and Least

Significant Bit Substitution. Transactions on Engineering and Computing Sciences, 11(3).122-134.

URL: http://dx.doi.org/10.14738/tecs.113.14919.

[Phase I]

Step 3. If 2τ ≤ |η| then read in next two bits o ∈ {00, 01, 10, 11} from W and embed  and o

in xi and xi+1, respectively, via the LSB technique. Proceed to Step 8.

Step 4. If the conditions of  = α and 0 ≤ (xi − xi+1) < τ are satisfied, then the bundle “carried”

α and go to Step 8, otherwise, if  = α, then repeat adjust both coefficients xi and xi+1 either by

increment or decrement, respectively, until 0 ≤ (xi − xi+1) < τ is true and proceed to Step 8.

Step 5. If the conditions of  = β and − τ ≤ (xi− xi+1) < 0 are satisfied, then the bundle “carried”

β and go to Step 8, otherwise, if  = β, then repeat adjust both coefficients xi and xi+1 either by

increment or decrement, respectively, until −τ ≤ (xi − xi+1) < 0 is true and proceed to Step 8.

Step 6. If the conditions of  = γ and τ ≤ (xi − xi+1) < 2τ are satisfied, then the bundle “carried”

γ and go to Step 8, otherwise, if  = γ, then repeat adjust both coefficients xi and xi+1 either by

increment or decrement, respectively, until τ ≤ (xi − xi+1) < 2τ is true and proceed to Step 8.

Step 7. If the conditions of  = δ and − 2τ ≤ (xi− xi+1) < τ are satisfied, then the bundle “carried”

δ and go to Step 8, otherwise, repeat adjust both coefficients xi and xi+1 either by increment or

decrement, respectively, until −2τ ≤ (xi − xi+1) < −τ is true.

[Phase II]

Step 8. Read in next two data bits ν ε{00,01,10,11} from W and compute the offset κ =

(xi

-xi+1)

2

-xi+2

Step 9. If 2τ ≤ |κ|is true, then embed ν in xi+2 via the LSB technique. Return to Step 1.

Step 10. If the conditions of ν = α and 0 ≤ (

(xi

-xi+1)

2

-xi+2) < τ are satisfied, then the bundle

“carried” α and go to Step 1, otherwise, if ν = α, then repeat adjust the coefficients (xi, xi+1) and

xi+2 either by increment or decrement, respectively, until 0 ≤ (

(xi

-xi+1)

2

-xi+2) < τ is true and

proceed to Step 1.

Step 11. If the conditions of ν = β and -τ ≤ (

(xi

-xi+1)

2

-xi+2) < 0 are satisfied, then the bundle

“carried” α and go to Step 1, otherwise, if ν = β, then repeat adjust the coefficients (xi, xi+1) and

xi+2 either by increment or decrement, respectively, until -τ ≤ (

(xi

-xi+1)

2

-xi+2) < 0 is true and

proceed to Step 1.

Step 12. If the conditions of ν = γ and τ ≤ (

(xi

-xi+1)

2

-xi+2) < 2τ are satisfied, then the bundle

“carried” γ and go to Step 1, otherwise, if ν = γ, then repeat adjust the coefficients (xi, xi+1) and

xi+2 either by increment or decrement, respectively, until τ ≤ (

(xi

-xi+1)

2

-xi+2) < 2τ is true and

proceed to Step 1.

Page 5 of 13

126

Transactions on Engineering and Computing Sciences (TECS) Vol 11, Issue 3, June - 2023

Services for Science and Education – United Kingdom

Step 13. If the conditions of ν = δ and -2τ ≤ (

(xi

-xi+1)

2

-xi+2) < -τ are satisfied, then the bundle

“carried” δ and go to Step 1, otherwise, if ν = δ, then repeat adjust the coefficients (xi, xi+1) and

xi+2 either by increment or decrement, respectively, until-2τ ≤ (

(xi

-xi+1)

2

-xi+2) < -τ is true and

proceed to Step 1.

Step 14. Stop.

Bit Extraction

The procedure of bit extraction is much simpler than that of bit embedding for the proposed

ECAT with the LSB method. The major steps of bit extraction are summarized in the following

algorithm.

Algorithm 2: Extracting hidden message from marked ECG.

Input: Marked ECG Ĥand control integer τ.

Output: Extracted message W.

Method:

Step 1. Input a marked bundle {x̂i

}i=0

n-1

from Ĥ and compute the offset η = x̂i

-x̂i+1. If the end of

input is encountered, then go to Step 6.

[Phase I]

Step 2. If |η| < 2τ is true, then two hidden bits can be extracted from the bundle, otherwise

proceed to Step 3. That is, if 0 ≤ (x̂i

-x̂i+1) < τ, then “00” are recognized, else if -τ ≤ (x̂i

-x̂i+1) <

0, then “01” is identified, else if τ ≤ (x̂i

-x̂i+1) < 2τ, then “10” are recognized, otherwise, “11” is

identified.

Proceed to Step 4.

Step 3. Four hidden bits can be identified from the coefficients x̂i and x̂i+1 via the LSB.

[Phase II]

Step 4. Compute the offset κ =

x̂i

-x̂i+1

2

-x̂i+2 If the condition of |κ| ⋖ 2τ is satisfied, then two 2

hidden bits can be extracted from the bundle, otherwise proceed to Step 5. Namely, data bits

“00,” “01,” “10,” and “11,” can be recognized if one of the following conditions:

0 ≤ (

x̂i

-x̂i+1

2

-x̂i+2) < τ,-τ ≤ (

x̂i

-x̂i+1

2

-x̂i+2) < 0, τ ≤ (

x̂i

-x̂i+1

2

-x̂i+2) < 2τ, and -2τ ≤ (

x̂i

-x̂i+1

2

-x̂i+2) <

-τ is satisfied, respectively. Return to Step 1.

Step 5. Two hidden bits can be identified from the coefficient x̂i+2 via the LSB. Proceed to Step

1.

Step 6. Assemble all of the extracted bits and form a watermark.

Page 6 of 13

127

Ching-Yu, Y., Tsung-Hsiang, H., & Pei-Yun, C. (2023). A New ECG Steganography Based on Enhanced Coefficient Adjustment Technique and Least

Significant Bit Substitution. Transactions on Engineering and Computing Sciences, 11(3).122-134.

URL: http://dx.doi.org/10.14738/tecs.113.14919.

Step 7. Stop.

From the above Algorithms we can see that the control parameter  plays an important role in

the proposed ECAT method. In fact, the SNR performance of the algorithm is gradually

decreased as  increased. The reason is that the larger the value of , the larger the number of

bundles encoded by the ECAT, which may result in significant distortion from the marked

bundle. On the other hand, the payload size of the algorithm is increased as  decreased. Due to

the less value of , the larger the number of bundles encoded by LSB, which provides the larger

hiding space.

EXPERIMENTAL RESULTS

A total of 23-set of ECG hosts were tested in our simulations. These test data were obtained

from the MIT-BIH Arrhythmia Database [14]. Each ECG host was composed of 30,000 samples.

Two commonly used objective measures such as signal-to-noise ratio (SNR) and percentage

residual difference (PRD) are used to evaluate performance. Both are defined as follows:

SNR = 10 log10

∑ si

2

i

∑ (si

-ŝi

)

2

i

(3)

and

PRD = √

∑ (si

-ŝi

)

2

i

∑ si

2

i

100% (4)

where si

, and ŝistand for the coefficients in the original ECG, marked ECG, respectively. First,

the payload/SNR/PRD performance of traditional CAT [13] is compared with that of our ECAT.

Table 1 indicated that the average payload of our method is 1.8 times larger than that of

traditional CAT [13] with the competitive SNR and PRD values. It can be summarized from

Table 1 that the ECAT (using  = 3) do provide larger hiding storage than traditional CAT (using

= 20) does. Subsequently, trade-off between SNR and payload for the proposed ECAT with the

LSB method using various  (from 2 to 9) is depicted in Fig. 1.

At first glance, it seems that the performance cures in the figure against common sense. (Since

the larger the payload size, the less the SNR value.) As mentioned in Section 2.2, the less 

employed by the proposed method, the less number of the host blocks encoded by ECAT, and

the larger number of the blocks encoded by LSB. Namely, it resulted in less distortion

introduced from the procedure of the coefficient adjustments, but increased hiding storage by

the LSB substitution. From the figure we can see that ECG106 has the best performance in

payload and SNR among the test ECGs, followed by ECG104, and ECG100 is the worst. In

addition, relationship between PRD and payload of our method is depicted in Fig. 2. The figure

confirmed the fact that the larger the SNR, the less the PRD, and vice versa. As we know, the

larger the SNR, the smaller the PRD, and the better performance obtained by an ECG

steganography.

Page 7 of 13

128

Transactions on Engineering and Computing Sciences (TECS) Vol 11, Issue 3, June - 2023

Services for Science and Education – United Kingdom

Table 1. Performance comparison between traditional CAT and our ECAT.

ECG Payload/SNR/PRD

hosts Traditional CAT†

[13] Our ECAT

100 19,135/ 54.68/ 0.184 36,822 / 52.03/ 0.250

101 19,142/ 53.17/ 0.220 35,362 / 51.80/ 0.257

102 19,452/ 53.07/ 0.222 36,160/ 51.71/ 0.260

103 18,975/ 53.33/ 0.190 35,396 / 51.81/ 0.257

104 18,931/ 51.55/ 0.264 33,024 / 52.03/ 0.250

105 18,591/ 52.21/ 0.245 33,174 / 51.86/ 0.255

106 19,125/ 52.17/ 0.246 33,692 / 51.60/ 0.263

108 19,786/ 51.37/ 0.270 34,198/ 51.62/ 0.262

109 18,107/ 51.70/ 0.260 31,788/ 51.83/ 0.256

Average 19,027/ 52.58/ 0.233 34,402/ 51.81/ 0.257

†When the method used the size of the host block is 3.

Fig. 1. Trade-off between SNR and payload of our method.

Page 8 of 13

129

Ching-Yu, Y., Tsung-Hsiang, H., & Pei-Yun, C. (2023). A New ECG Steganography Based on Enhanced Coefficient Adjustment Technique and Least

Significant Bit Substitution. Transactions on Engineering and Computing Sciences, 11(3).122-134.

URL: http://dx.doi.org/10.14738/tecs.113.14919.

Fig. 2. Trade-off between PRD and payload of our method.

To further examine the hiding performance of the proposed method, the resulting payload and

SNR of the test ECGs with five different  are illustrated in Fig. 3. Figure 3(a) shows that the

hiding capacity provided by ECG107 has the largest payload size, followed by ECG109, and ECG

108 provided the least payload as  < 2. Figure 3(b) indicates that ECG 107 provided the best

SNR as  < 4 and the least SNR as  > 2. The distortions introduced by the proposed method

using  = 6 (in the first 5-s interval) from the marked ECGs are illustrated in Fig. 4. The marked

signal introduced by our method (red line) was approximately similar to the original one (blue

line).

(a)

Page 9 of 13

130

Transactions on Engineering and Computing Sciences (TECS) Vol 11, Issue 3, June - 2023

Services for Science and Education – United Kingdom

(b)

Fig. 3. Hiding performance of our method using five different . (a) payload and (b) SNR.

(a)

Page 10 of 13

131

Ching-Yu, Y., Tsung-Hsiang, H., & Pei-Yun, C. (2023). A New ECG Steganography Based on Enhanced Coefficient Adjustment Technique and Least

Significant Bit Substitution. Transactions on Engineering and Computing Sciences, 11(3).122-134.

URL: http://dx.doi.org/10.14738/tecs.113.14919.

(b)

(c)

(d)

Page 11 of 13

132

Transactions on Engineering and Computing Sciences (TECS) Vol 11, Issue 3, June - 2023

Services for Science and Education – United Kingdom

(e)

(f)

Fig. 4. Close observations of the marked ECGs introduced by our method using  = 6. (a)

ECG100, (b) ECG101, (c) ECG102, (d) ECG111, (e) ECG112, and (f) ECG113.

Performance comparison of existed techniques and our method in several test ECGs is given in

Table 2. It is clear that the payload of our method is the largest among the compared methods.

Namely, the payload provided by our method is two times larger than that provided by Yang

and Wang [10], and is nearly two times bigger than that provided by Yang et al. [12]. In addition,

the hiding storage of the proposed method is 1.84 times larger than that of the Yang and Wang

[13]. In summary, the averages SNR of the proposed method is superior to that of other three

CAT-based schemes. Furthermore, another objective measure which often used to evaluate the

performance of data hiding is peak signalto-noise ratio (PSNR). The PSNR and payload

performance comparison of various methods is listed in Table 3. Obviously, the average

payload of our method is 6.13 times larger than that of the other two schemes [8, 11], while our

PSNR is still better than theirs. The PSNR is defined as follows:

Page 12 of 13

133

Ching-Yu, Y., Tsung-Hsiang, H., & Pei-Yun, C. (2023). A New ECG Steganography Based on Enhanced Coefficient Adjustment Technique and Least

Significant Bit Substitution. Transactions on Engineering and Computing Sciences, 11(3).122-134.

URL: http://dx.doi.org/10.14738/tecs.113.14919.

PSNR = 10 log10

Max(si

)

2

1

K

∑ (si

-ŝi

)

2

i

(5)

where Max(

s

i) and K stand for the maximum value of si and the number of input samples,

respectively. From Tables 2 and 3 we can conclude that an efficient ECG steganography is

achieved by the proposed method.

Table 2. SNR/payload performance comparison between various methods.

ECG data SNR/Payload

Yang and Wang

[10]

Yang et al.* [12] Yang and Wang†

[13]

Our method

100 48.19/ 19,604 50.96/ 21,349 50.18/ 24,146 50.36/ 41,156

101 47.39/ 19,610 48.17/ 21,911 49.07/ 24,427 50.00/ 41,356

102 51.60/ 19,602 49.74/ 21,702 50.04/ 23,542 50.04/ 41,222

103 44.92/ 19,600 49.53/ 21,171 49.93/ 21,473 49.89/ 41,459

104 46.84/ 19,602 47.61/ 21,265 49.95/ 21,312 50.17/ 42,304

111 49.43/ 19,630 47.49/ 21,638 50.28/ 21,651 49.83/ 41,894

200 47.36/ 19,608 47.53/ 20,808 50.40/ 20,474 50.01/ 42,780

201 50.13/ 19,612 45.94/ 21,153 50.40/ 23,160 50.51/ 41,506

202 50.16/ 19,632 49.01/ 21,407 50.46/ 23,599 49.69/ 41,096

232 51.58/ 19,620 47.67/ 21,846 49.08/ 22,542 49.57/ 41,128

Average 48.76/ 19,612 48.37/ 21,425 49.95/ 22,633 50.01/ 41,590

*The size of a host block used by the method is 4*4

† The size of the host blocks employed by the method is larger than 3 * 3.

*Our method used the control parameter  = 4.

Table 3. PSNR/payload performance comparison of various methods.

Jero et al. [8] Pandey et al. [11] Our method

54.75/ 10,650 58.01/ 10,650 63.01/ 9,800

51.13/ 14,450 55.55/ 14,450 61.83/ 12,800

45.12/ 18,022 53.62/ 18,022 59.84/ 20,000

39.52/ 21,873 51.98/ 21,873 58.29/ 28,800

34.46/ 25,149 50.79/ 25,149 56.86/ 39,200

Average 45.00/ 18,029 Average 53.99/ 18,029 Average 59.97/ 110,600

CONCLUSION

In this article, we presented an efficient ECAT to promote the payload size of traditional CAT

scheme. The payload of the former is apparently larger than that of the latter, while the

resultant SNR (or perceived quality) of ECAT is competitive to that of CAT. In addition, a large

number of data bits can be embedded in an ECG host as the proposed ECAT with the LSB

method. Experimental results confirmed that both the average payload and SNR/PSNR of our

method are superior to those of the variants of the CAT and existed schemes. In addition, the

resultant perceived quality is good. Due to the algorithm is simple, the application of our

method can be found in portable (or wearable) biometric measurements.

Page 13 of 13

134

Transactions on Engineering and Computing Sciences (TECS) Vol 11, Issue 3, June - 2023

Services for Science and Education – United Kingdom

References

[1]. Eielinska, E., et al. Trends in steganography. Comm. of the ACM, 2014. 57(3): p. 86-95, doi:

10.1145/2566590.2566610.

[2]. Hussain, M. A., et al., Image steganography in spatial domain: A survey, Signal Processing: Image

Communication, 2018. 65(7): p. 46-66, DOI: 10.1016/j.image.2018.03.012.

[3]. Hsiao, C.Y., et al., Simple and robust watermarking scheme based on square-root-modulus technique,

Multimedia Tools and Applications, 2018. 77(5): p. 30419-30435, DOI: 10.1007/s11042-018-6121-3.

[4]. Liu, S., et al., Digital image watermarking method based on DCT and fractal encoding, IET Image Proc., 2017.

11(10): p. 815-821, DOI: 10.1049/iet-ipr.2016.0862.

[5]. Sahu, A.K. and Swain, G., Data hiding using adaptive LSB and PVD technique resisting PDH and RS analysis,

Int. Journal of Electronic Security and Digital Forensics, 2019. 11(6): p. 458-476, DOI:

10.1504/IJESDF.2019.102567.

[6]. Kadhim, I.J., et al., Comprehensive survey of image steganography: Techniques, Evaluations, and trends in

future research, Neurocomputing, 2019. 335(3): p. 299-326, DOI: 10.1016/j.neucom.2018.06.075.

[7]. Yang, C.Y. and Wang, W.F., Effective electrocardiogram steganography based on coefficient alignment,

Journal of Medical Sys., 2016. 40(12), DOI: 10.1007/s10916-015-0426-9.

[8]. Jero, S.E., et al., Imperceptability-robustness tradeoff studies for ECG steganography using continuous ant

colony optimization, Expert Systems with Applications, 2016. 49(1): p. 123-135, DOI:

10.1016/j.eswa.2015.12.010.

[9]. Shiu, H.J., et al., Preserving privacy of online digital physiological signals using blind and reversible

steganography, Computer Methods and Programs in Biomedicine, 2017. 151(11): p. 159-170, DOI:

10.1016/j.cmpb.2017.08.015.

[10]. C.Y. Yang and W.F. Wang, “An improved high-capacity ECG steganography with smart offset coefficients,”

The 14th Int. Conf. on Intelligent Information Hiding and Multimedia Signal Processing (IIH-MSP 2018),

Nov. 26-28, Sendai, Japan, 2018.

[11]. Pandey, A., et al., A novel fused coupled chaotic map based confidential data embedding-then-encryption of

electrocardiogram signal, Biocybernetics and Biomedical Engineering, 2019. 39(2): p. 282-300, DOI:

10.1016/j.bbe.2018.11.012.

[12]. Yang, C.Y., et al., Adaptive electrocardiogram steganography based on 2D approach with predetermined

rules, Asian Journal of Computer and Information Systems, 2020. 8(1): p. 1-10, DOI: 10.24203/ajcis.

v8i1.6059.

[13]. Yang, C.Y. and Wang, W.F., Progressive data hiding in integer wavelet transform of electrocardiogram by

using simple decision rule and coefficient calibration, Revue d'Intelligence Artificielle, 2020. 34(2): p. 1120,

DOI: 10.18280/ria.340102.

[14]. Moody, G.B. and Mark, R.G. The impact of the MIT-BIH arrhythmia database. IEEE Eng. in Medical and Biol.,

2001. 20(3): p. 45-50, DOI: 10.1109/51.932724.