In this section, we introduce two extended types (Types 1 and 2) of image sharing; we also demonstrate how they can be implemented by using either the polynomial approach or linear-equations approach.
To understand what Type 1 image sharing is, without loss of generality, consider 6 people {P1,…, P6} who share together a secret of the company. The secret is such that 3 or more people must be gathered to unveil it. Hence, the secret can be recovered by () = 20 possible combinations of personnel if no combination is forbidden. However, according to a security check, P1, P3, and P5 worked for a rival company before working in ours, making them less trustworthy. The combination of these 3 employees should thus be excluded as a combination that allows access to the secret. This is an example of Type 1 image sharing with a single forbidden combination, namely, {P1, P3, P5}. Of course, if the boss of our company wishes to be more careful, he can even forbid more combinations such as {P1, P3, x}, {P1, P5, x}, and/or {P3, P5, x}, where x can be any of the other employees (i.e., any other Pi). In the preceding example, some combinations cannot be used to gain access to the secret. We call it “sharing with forbidden combinations” and denote it as (tparticipants, nparticipants, f) sharing. Here,
For example, the preceding example is a (3, 6, 1) scheme if there is only one forbidden combination {P1, P3, P5} and a (3, 6, 4) scheme if all {P1, P3, x} combinations are forbidden. Traditional sharing, which has no forbidden combinations, can be denoted as (tparticipants, nparticipants, f) = (3, 6, 0), and it is thus treated as a special case of Type 1 image sharing.
3.1. Using the Polynomial Approach to Design Types 1 and 2
Type 1 (Sharing with forbidden combinations):
Without loss of generality, consider 6 participants {
P1,…,
P6} in the company, of which, any 3 can unveil the secret, unless the 3 participants constitute a forbidden combination. Examples 1 and 2 illustrate the steps in creating a sharing scheme with one and two forbidden combinations, respectively, and the
Appendix A at the end of paper illustrates the cases when the number of forbidden combinations is 3 or 4. In general, the design of Type 1 is case by case, and the design of Type 2 is easier.
Example 1 (one forbidden combination). A sharing scheme with one forbidden combination is easy to design. Without loss of generality, let the only forbidden combination be the combination {P4, P5, P6}. Because there are only 6 participants, we shall create 6 “shadows” for the 6 participants per the following steps 1 and 2. This results in the creation of a (tparticipants = 3, nparticipants = 6, f = 1) threshold scheme, where {P4, P5, P6} is the only forbidden combination.
Step 1: Use (tartificial, nartificial) = (9, 17) in polynomial-based sharing to share the secret and obtain nartificial = 17 shares {S1,…, S17} such that any tartificial = 9 of them can recover the secret.
Step 2: Distribute these 17 shares to 6 participants. The distributions of the 17 shares are P1 = {S1, S2, S3}, P2 = {S4, S5, S6}, P3 = {S7, S8, S9}, P4 = {S10, S11, S12}, P5 = {S13, S14, S15}, and P6 = {S16, S17, S12, S15}.
Note that we deliberately let the final two components of the participant P6 be the already-used shares S12 and S15, which had already appeared in P4 and P5. By doing this, the total number of different shares in the combination {P4, P5, P6} is only 3 + 3 + 2 = 8, which is less than the threshold of 9, meaning that the secret cannot be revealed. Conversely, any other combination {Pi, Pj, Pk} yields at least 9 distinct shares, and hence, can reveal the secret. Notably, for Participants P1 to P5, every participant holds a shadow (the data held by a participant), which is constituted by 3 shares; and each share is 9 times smaller in size than the original secret in the (9, 17) threshold scheme. Therefore, in our (tparticipants = 3, nparticipants = 6, f = 1) scheme, the data size of each participant P1 to P5 is 3 × 1/9 = 1/3 of the size of the original secret. By contrast, because P6 holds 4 shares, the data size is 4 × 1/9 = 4/9 of the size of the original secret.
Example 2 (
two forbidden combinations)
. Figure 1 illustrates the design. Still assume
tparticipants = 3, and
nparticipants = 6. Hence, any 3 participants can unveil the secret, unless the 3 participants constitute a forbidden combination. In Example 2, only two combinations {
P1,
P2,
P3} and {
P4,
P5,
P6} are forbidden, hence,
f = 2. These combinations may be forbidden because, for example, the groups {
P1,
P2,
P3} and {
P4,
P5,
P6} do not trust each other. Because we have 6 participants, we now create 6 shadows for these 6 participants, where each participant gets a shadow comprising some shares. As with the first step of Example 1, we still use
tartificial = 9 in the traditional polynomial sharing scheme to share the secret and obtain many shares {
S1,
S2,
…,
S13,
…}, allowing the secret to be unveiled if any 9 of the shares are used. Because the threshold for “number of shares” is 9 and
P1 = {
S1,
S2,
S3},
P2 = {
S4,
S5,
S6}, and
P3 = {
S7,
S8,
S3,
S6}, the combination {
P1,
P2,
P3} yields only 8 shares {S
1,…, S
8}, which is still fewer than 9, and the secret cannot be unveiled. Likewise, {
P4,
P5,
P6} yields only 8 shares {
S9,…,
S16} because
P4 = {
S9,
S10,
S11},
P5 = {
S12,
S13,
S14}, and
P6 = {
S15,
S16,
S11,
S14}. Thus, the secret cannot be unveiled under the combination {
P4,
P5,
P6}. As for the other combinations {
Pi,
Pj,
Pk}, members in the combination can view the secret because at least 9 shares are available. The proof of the aforementioned claims is routine and thus not presented for brevity.
Table 1 presents some other examples.
Notably, because more constraints must be considered, the design is likely to become more difficult when the number of forbidden combinations increases.
Example 3 (
3, nparticipants, f)
. Sharing whose f forbidden combinations always contain a trouble-making couple P1 and P2. See the Appendix A for the design. Type 2 (Sharing with cross-department support):
Let each participant belong to one of several departments, for example, 3 departments {
DEPH,
DEPM,
DEPL}. Without loss of generality, assume the parameter pair (
t = threshold,
n = number of people in this department) for the 3 departments are, respectively, (
tH = 3,
nH = 4) for
DEPH, (
tM = 5,
nM = 6) for
DEPM, and (
tL = 7,
nL = 7) for
DEPL. As the first step in the design, we shall assign a value to an artificial threshold
tartificial, which is larger than any given local threshold. Subsequently, we use this artificial threshold
tartificial to share the secret and thus obtain several shares such that any
tartificial shares can recover the secret. We then distribute these shares to each person of each department. In this cross-department support type, we assume that shares are not repeatedly distributed. Let each person in
DEPH,
DEPM, and
DEPL obtain, respectively,
QH,
QM, and
QL shares. Then, to satisfy each department’s rule for unveiling the secret, we must have:
Because tartificial, QH, QM, and QL are all positive integers, if QL is 1, then Equation (2) implies 7 ≥ tartificial > 6; hence, tartificial = 7. Furthermore, the right-hand side of Equation (3) implies tartificial = 7 > 4 × QM, which means the positive integer QM is 1, thus contradicting the left-hand side of Equation (3) because 5 × QM ≥ tartificial then becomes 5 × 1 ≥ 7. Hence, we cannot use QL = 1. Thus, we try the next smallest value 2 for QL. Equation (2) then implies 14 ≥ tartificial > 12. Hence, tartificial is 13 or 14, proven as follows. If tartificial = 13, Equation (3) becomes 5QM ≥ 13 > 4QM, which can be satisfied by QM = ⌈13/5⌉ = 3, and Equation (4) becomes 3QH ≥ 13 > 2QH, which can be satisfied by QH = ⌈13/3⌉ = 5. Similarly, if tartificial = 14, Equation (3) becomes 5QM ≥ 14 > 4QM, which can be satisfied by QM = ⌈14/5⌉ = 3, and Equation (4) becomes 3QH ≥ 14 > 2QH, which can be satisfied by QH = ⌈14/3⌉ = 5.
Likewise, if we try the next smallest value for QL, 3, then Equation (2) implies 21 ≥ tartificial > 18. Hence, tartificial is in {19, 20, 21}. We now prove that tartificial can be 19, 20, or 21. If tartificial = 19, Equation (3) becomes 5QM ≥ 19 > 4QM, which can be satisfied by QM = ⌈19/5⌉ = 4, and Equation (4) becomes 3QH ≥ 19 > 2QH, which can be satisfied by QH = ⌈19/3⌉ = 7. Analogous steps can be taken to prove that tartificial can be 20 or 21.
When we execute the preceding procedure for QL ≥ 2, we obtain valid values for tartificial, namely, tartificial ∈ {13, 14}when QL = 2, {19, 20, 21}when QL = 3, {25, 26, 27, 28}when QL = 4, {31, 32, 33, 34, 35}when QL = 5, or {37, 38, 39, 40, 41, 42}when QL = 6, and so on.
We now examine the system further. For example, let the artificial threshold be 14. We can then use (
tartificial,
nartificial) = (14, 52) in the traditional polynomial-based sharing scheme to create 52 shares. Each participant in
DEPH,
DEPM, and
DEPL gets 5 = ⌈14/3⌉, 3 = ⌈14/5⌉, and 2 = ⌈14/7⌉ shares, respectively. Moreover, for any 2 participants (whether from the same department or from different departments), their set of shares do not intersect. Notably, because no intersection is allowed, any 3
DEPH participants can hand in 3 × 5 = 15 > 14 distinct shares; any 5
DEPM participants can hand in 5 × 3 = 15 > 14 distinct shares; and any 7
DEPL participants can hand in 7 × 2 = 14 distinct shares. However, 2
DEPH participants receive only 2 × 5 = 10 < 14 distinct shares; 4
DEPM participants receive only 4 × 3 = 12 < 14 distinct shares; and 6
DEPL participants receive only 6 × 2 = 12 < 14 distinct shares. Hence, the intra-department thresholds to unveil the secret are 3 for
DEPH, 5 for
DEPM, and 7 for
DEPL. We create 52 shares because 4 × 5 + 6 × 3 + 7 × 2 = 52. In the deciphering meeting, the secret is revealed once the total number of available shares is equal to or greater than the threshold number 14, regardless of whether the participants are from the same department.
Table 2 lists some of the many possible combinations that can be used to reveal the secret.
For the 3 departments, the size of each shadow file held by a participant is, respectively, 5/14, 3/14, and 2/14 of the original secret image size. Moreover, in the aforementioned design for cross-department sharing, the artificial threshold
tartificial is not unique, and the designers have the freedom to choose their own artificial threshold
tartificial. For example, rather than using the aforementioned
tartificial = 14, we may also use, say,
tartificial = 20, to create shares. Then each participant in
DEPH,
DEPM and
DEPL has 7 (= ⌈20
/3⌉), 4 (= ⌈20
/5⌉) and 3(= ⌈20
/7⌉) shares, respectively. For each of the 3 departments, the shadow size is, respectively, 7/20, 4/20, and 3/20 of the size of the original secret. In addition to following the preceding steps to obtain valid values of
tartificial by checking whether Equations (2)–(4) are satisfied, another method for obtaining
tartificial is to use the least common multiple (LCM) of the threshold values of all departments. If the LCM is used, Equations (2)–(4) will be automatically satisfied: positive integer solutions for
QH,
QM, and
QL necessarily exist because we can let
QH =
tartificial/tH,
QM =
tartificial/tM, and
QL =
tartificial/tL. Algorithm 1 shows how to use LCM to create shadows for cross-department support.
Algorithm 1. To create shadows for cross-department support |
1: Set the value of tartificial to LCM (tH, tM, tL). 2: Set the value of nartificial to (tartificial/tH) × nH + (tartificial/tM) × nM + (tartificial/tL) × nL. 3: Use (tartificial, nartificial) in traditional polynomial sharing scheme to create nartificial shares. 4: for each participant in DEPH do 5: Grab (tartificial/tH) not-yet-used shares to create the shadow for this participant. 6: end for 7: for each participant in DEPM do 8: Grab (tartificial/tM) not-yet-used shares to create the shadow for this participant. 9: end for 10: for each participant in DEPL do 11: Grab (tartificial / tL) not-yet-used shares to create the shadow for this participant. 12: end for 13: return the shadow of each participant. |
Remark 1: Here, the shadow size of each participant in the 3 departments are, respectively, (
LCM/tH) × (1/
LCM) = 1
/tH, (
LCM/tM) × (1/
LCM) = 1
/tM, and (
LCM/tM) × (1/
LCM) = 1
/tL the size of the original secret. This is because, as stated in Thien and Lin [
4], each share is
LCM times smaller than the original secret is when
tartificial =
LCM is used to create shares.
Remark 2: Many values can be used as the value of tartificial, but they must satisfy the necessary but insufficient condition of tartificial ≥ Max{thresholds of all departments}. We may use multiple thresholds. For example, to confuse hackers, we can use tartificial = 19, 14, 13, 20, 21, 25, 26, 28, 27, 31, 32, 35, 34, 33, … to confuse hackers. For instance, a small part of the secret is shared using tartificial = 19; then a small part of the secret is shared using tartificial = 14; then certain parts of the secret are shared using tartificial = 13; then… Since there are so many possible combinations, it will be more difficult for the hackers.
Figure 2 illustrates the design (
Figure 2a–c) and the secret disclosure (
Figure 2d) for a cross-department support system involving 3 departments, as described in
Section 3.1. We use (
tartificial,
nartificial) = (14, 52) in traditional polynomial-based sharing to create 52 shares. Then, we distribute these 52 shares to all participants; and each participant of department
DEPH gets more shares to form their shadows.
3.2. Using the Linear-Equations Approach to Design Types 1 and 2
The linear-equations approach can be mapped from the polynomial approach. Because we already introduced the polynomial approach, we only need to know how to map from the polynomial approach to the linear-equations approach. In the polynomial approach, every participant has a shadow file comprising several shares, whereas in the linear-equations approach, every participant has a shadow file comprising several equations. For both approaches, the threshold
tartificial must be met to unveil the secret. Specifically, in the polynomial approach, participants attending a meeting must have at least
tartificial distinct shares, whereas in the linear-equations approach, participant attendees must have at least
tartificial independent equations. The linear-equations approach thus proceeds as follows.
Algorithm 2. General procedure to derive the linear-equations approach from the (tartificial, nartificial) polynomial approach |
1: Create an nartificial-by-tartificial matrix A so that any tartificial of its nartificial rows are independent. 2: Partition secret image to non-overlapping segments of tartificial pixel values each. 3: for each secret segment, Dsecret, do 4: for k = 1 to nartificial do 5: Calculate the inner product value IPk = Rowk ∙ Dsecret. 6: end for 7: for i = 1 to nparticipants do 8: for j = 1 to nartificial do 9: if Participant Pi gets the share Sj in the polynomial approach then 10: Participant Pi gets the pair (Rowj, IPj) in linear equations approach. 11: end if 12: end for 13: end for 14: end for 15: return the rows and inner product values collected by each participant. |
Notably, every share
Sj in the polynomial approach is replaced by a corresponding equation
Eqj in the linear-equations approach. Here, each equation
Eqj is formed of two parts: firstly, the
tartificial left-hand side coefficients of equation
Eqj are the
tartificial-dimensional vector
Rowj; secondly, the right-hand side of equation
Eqj is the inner product value
IPj. Hence, any
tartificial independent equations can uniquely solve for the
tartificial values of the
tartificial-dimensional vector
Dsecret. If a participant
P gets {
S1,
S2,
S5,
S7} in the polynomial approach, then that participant
P also gets {
Eq1,
Eq2,
Eq5,
Eq7}, i.e. {
Rowj &
IPj}
j = 1, 2, 5, 7 in the linear-equations approach. Thus, regardless of whether a Type 1 (the forbidden type) or Type 2 (the cross-department support type) system is used, just map the method from polynomial approach to linear-equations approach, by letting the shares held by each participant be replaced by the corresponding equations, or vice versa. For example, if the polynomial approach is used, where participant
Pa gets {
S1,
S2,
S5,
S7} and participant
Pb gets {
S3,
S4,
S6,
S8}, to obtain the corresponding instance of the linear-equations approach, we let participant
Pa get {
Rowj; IPj}
j = 1, 2, 5, 7 and participant
Pb get {
Rowj; IPj}
j = 3,4, 6, 8. Moreover, if rows are created using a specified process or algorithm, as discussed in
Section 4.1, then each participant does not need to store the rows; for example,
Pa only stores {
IPj}
j = 1, 2, 5, 7 and
Pb only stores {
IPj}
j = 3, 4, 6, 8.
Types 1 and 2 are detailed as follows. Example 1* is derived from Example 1 of
Section 3.1, where Example 1*’s steps are such that Step 1* is derived from Step 1 of
Section 3.1, Step 2* is derived from Step 2 of
Section 3.1, and so on.
Type 1: Sharing with forbidden combinations
Because the linear-equations approach can be mapped from the polynomial approach, we can, without loss of generality, simply use the examples in
Section 3.1 to detail such a mapping. Hence, as in
Section 3.1, we still consider 6 participants (
P1,…,
P6), where any 3 of the 6 participants can unveil the secret, unless the 3 participants form a forbidden combination. Examples 1* and 2* illustrate the steps required to create the sharing scheme with one and two forbidden combinations, respectively. Every other example of
Section 3.1 also has a linear-equations counterpart in this section,
Section 3.2.
Example 1* (
one forbidden combination)
. As in Example 1 of Section 3.1, {P4, P5, P6} is the only forbidden combination. Because there are 6 participants, we create 6 shadows for the 6 participants per the following steps 1* and 2*. Notably, this is a tparticipants = 3, nparticipants = 6, f = 1) threshold scheme, and {P4, P5, P6} is the only forbidden combination. Step 1*: As in Step 1 of Example 1 for the polynomial-based approach, tartificial = 9 and nartificial = 17. First, create a matrix with 17 rows where each row is 9-dimensional and any tartificial of the nartificial rows are independent. Subsequently, grab the next tartificial = 9 not-shared-yet numbers from the secret, where these secret numbers are termed Dsecret. Then, for i = 1 to i = nartificial (1 to 17 in this case), let Eqi be the equation (Rowi) ∙ (Dsecret) = IPi, where IPi is the value of the inner product of the two vectors Rowi and Dsecret. Notably, because any tartificial = 9 of the nartificial rows are independent, any tartificial = 9 of the nartificial equations can uniquely solve for Dsecret, which has tartificial = 9 secret numbers.
Step 2*: Distribute these 17 equations to 6 participants. Recall that in Example 1’s polynomial-based approach (
Section 3.1), the distribution of shares were
P1 = {
S1,
S2,
S3},
P2 = {
S4,
S5,
S6},
P3 = {
S7,
S8,
S9},
P4 = {
S10,
S11,
S12},
P5 = {
S13,
S14,
S15}, and
P6 = {
S16,
S17,
S12,
S15}. Thus, for this corresponding instance of the linear-equations approach, we let the distribution of equations for the 6 participants be
P1 = {
Eq1,
Eq2,
Eq3},
P2 = {
Eq4,
Eq5,
Eq6},
P3 = {
Eq7,
Eq8,
Eq9},
P4 = {
Eq10,
Eq11,
Eq12},
P5 = {
Eq13,
Eq14,
Eq15}, and
P6 = {
Eq16,
Eq17,
Eq12,
Eq15}.
Similarly, the final two components of the participant P6 already appeared in P4 and P5. Thus, the total number of different equations in the combination {P4, P5, P6} is only 3 + 3 + 2 = 8, which is less than the threshold tartificial = 9, meaning that the secret Dsecret cannot be revealed. Conversely, any other combination {Pi, Pj, Pk} yields 9 independent equations, meaning that the 9 secret numbers in Dsecret can be revealed.
Remark 3: Assume that the rule to create the independent 17-by-9 matrix A in Step 1* is from an algorithm—for example, each element ai,j is (i)j−1—then there is no need to store the matrix. In this case, only the right-hand side of that equation—for example, the value IPi of the inner product for Eqi—needs to be stored. Hence, when a 9-number Dsecret is shared, every participant from P1 to P5 holds 3 inner product values, and P6 holds 4 inner product values.
Example 2* (two forbidden combinations). Similarly, assume that tparticipants = 3 and nparticipants = 6, meaning that any 3 participants can unveil the secret, unless the 3 participants constitute a forbidden combination. As in Example 2 of the polynomial approach, only the two combinations {P1, P2, P3} and {P4, P5, P6} are forbidden; hence, f = 2. Because we have 6 participants, we create 6 shadow files for these 6 participants. Each participant gets a shadow comprising some equations. As in the first step of Example 1*, we use the (tartificial = 9, nartificial = 16) scheme to share the 9-number secret section Dsecret and obtain 16 equations {Eq1, …, Eq16}, where the 9-number secret section Dsecret can be obtained if any 9 equations are used. Because the threshold for “number of equations” is 9 and P1 = {Eq1, Eq2, Eq3}, P2 = {Eq4, Eq5, Eq6}, and P3 = {Eq7, Eq8, Eq3, Eq6}, we can see that {P1, P2, P3} 3 people together only have 8 equations {Eq1,…, Eq8}, which is still less than 9, meaning that Dsecret cannot be unveiled. Similarly, {P4, P5, P6} together only have 8 equations {Eq9,…, Eq16} because P4 = {Eq9, Eq10, Eq11}, P5 = {Eq12, Eq13, Eq14}, and P6 = {Eq15, Eq16, Eq11, Eq14}, meaning that Dsecret also cannot be unveiled. As for the other {Pi, Pj, Pk} combinations, the secret can be revealed because at least 9 equations are available. The proofs of the preceding statements are routine and are thus not presented.
Type 2: Cross-department support system
We must demonstrate how the polynomial approach can be extended to the linear-equations approach. Without loss of generality, only demonstrate such an extension for the example where
tartificial = 14 in the cross-department support algorithm (Algorithm 1) of
Section 3.1. Other examples of Type 2 in
Section 3.1 can be analogously extended. In Algorithm 3, still use the parameter values used in
Figure 2, namely, (
tH = 3,
nH = 4), (
tM = 5,
nM = 6), (
tL = 7,
nL = 7), and
tartificial = 14.
Algorithm 3. Linear-equations approach to cross-department support |
1: Get the input thresholds {tH = 3, tM = 5, tL = 7} of all 3 departments. Also get the input number of participants {nH = 4, nM = 6, nL = 7} of all 3 departments. 2: Use Equations (2)–(4) to find a suitable artificial threshold value tartificial =14. 3: Calculate nartificial = ⌈tartificial/tH⌉ × nH + ⌈tartificial/tM⌉ × nM + ⌈tartificial/tL⌉ × nL = ⌈14/3⌉ × 4 + ⌈14/5⌉ × 6 + ⌈14/7⌉ × 7 = 5 × 4 + 3 × 6 + 2 × 7 = 52. 4: Create a 52-by-14 matrix A such that any 14 of the 52 rows are independent. 5: Partition the secret image to non-overlapping segments of 14 pixel values each. Let nSEG denote the total number of non-overlapping segments. 6: While there are secret segments not shared yet do 7: Grab a tartificial-values not-yet-shared segment, Dsecret, of the secret image. 8: for k = 1 to nartificial do 9: Calculate the inner product value IPk = Rowk ∙ Dsecret for this segment. 10: end for 11: end while 12: for each participant in DEPH do 13: Grab ⌈tartificial/tH⌉ = ⌈14/3⌉ = 5 of the not-yet-assigned rows of A. The participant stores these 5 rows and the 5 × nSEG inner product values created above using these 5 rows. 14: end for 15: for each participant in DEPM do 16: Grab ⌈tartificial/tH⌉ = ⌈14/5⌉ = 3 of the not-yet-assigned rows of A. The participant stores these 3 rows and the 3 × nSEG inner product values created above using these 3 rows. 17: end for 18: for each participant in DEPL do 19: Grab ⌈14/7⌉ = 2 of the not-yet-assigned rows of A. The participant stores these 2 rows and the 2 × nSEG inner product values created above using these 2 rows. 20: end for 21: return the equation shadows, which are the stored rows and corresponding stored inner product values, of each participant of each department. |
Notably, in Algorithm 3, because any
tartificial = 14 of the
nartificial = 52 rows are independent, we can claim that any
tartificial = 14 of the
nartificial = 52 equations can solve for the secret
Dsecret uniquely [
21,
22], true for the secret
Dsecret of each segment. Again, if the rule to create the independent 52-by-14 matrix
A is by a predetermined method or rule (see
Section 4.1), then only the value
IPi of the inner product needs be stored for the equation
Eqi. Hence, when a 14-number segment
Dsecret is shared, every
DEPH,
DEPM, and
DEPL participant holds
⌈tartificial/tH⌉ =
⌈14/3
⌉ = 5,
⌈tartificial/tM⌉ =
⌈14/5
⌉ = 3, and
⌈tartificial/tL⌉ =
⌈14/7
⌉ = 2 inner product values, respectively. The 5:3:2 ratio is identical to the 5:3:2 ratio when the polynomial approach is used. Finally, as stated in
Section 3.1,
tartificial can have many possible values. Hence, to confuse hackers, we may use a sequence of dynamic thresholds such as
tartificial = 19, 14, 13, 20, 21, 25, 26, 28, 27, 31, 32, 35, 34, 33,… to share a single secret image.