GROUPING OF TRAJECTORIES IN THE ENCRYPTED DOMAIN
Patent Information
- Application Number
- DE602021021428
- Authority / Receiving Office
- DE · DE
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2020-05-29
- Filing Date
- 2021-05-21
- Publication Date
- 2024-11-06
- Estimated Expiration
- 2041-05-21
AI Technical Summary
Existing methods for comparing geographical trajectories, such as those recorded by GPS navigation assistants, fail to protect user identity effectively, as anonymization techniques do not sufficiently obscure sensitive information that can reveal personal details like residence or age from the data.
A method involving the determination of a signature vector through the application of a hash function to an encrypted vector, which filters out irrelevant components and allows for efficient comparison of trajectories while maintaining user anonymity by only considering transitions between geographic cells, thereby reducing computational complexity and protecting user identity.
This approach enables efficient comparison of geographical trajectories while ensuring significant protection of user identity by filtering out non-relevant components, reducing the complexity of the comparison process while maintaining precision, and preventing identification of individuals from their trajectories.
Description
[0001] The present invention relates to the field of analysis and computer processing of data relating to geographical trajectories, for example, the processing of data collected by GPS navigation assistants.
[0002] It finds applications in particular in computer systems comparing this data relating to these geographical trajectories.
[0003] Indeed, GPS navigation assistants may be required to compare the trajectories of their users in order to adapt the recommended routes.
[0004] In addition, the analysis of data relating to geographical trajectories, by making it possible to determine the most frequently used routes, offers the possibility of adapting these routes, for example, the signage or even the services offered on these routes.
[0005] However, data relating to geographical trajectories is sensitive data, as it very often makes it possible to identify the individuals who have traveled these trajectories. Respecting the "privacy" of users of an IT service is an increasingly strategic issue, it is important to be able to analyze data relating to geographical trajectories without compromising the identity of these users.
[0006] Some solutions rely on trajectories anonymization. Thus, a given trajectory is no longer associated with any element that allows the user to be identified.
[0007] However, a trajectory may contain sufficient information in itself to identify a user. Indeed, a trajectory of a user, in particular when it relates to the movement of a user over several days, may make it possible to determine significant locations in the user's life, for example, their place of work or their place of residence. A trajectory of a user, in that it makes it possible to know the speed of the user, may also make it possible to approximate the size or age of the user. Thus, the anonymization of trajectories does not offer sufficient protection of the identity of users of a service using data relating to their geographical trajectories. EP3640946A1 discloses a method for selecting target cells from a plurality of candidate cells, in the context of a therapeutic process, for example for the manufacture of a drug.
[0008] The publication "Incentive Mechanism Design for Crowdsourcing-based Indoor Localization", IEEE ACCESS, Li Wei and AI. discloses models of crowdsourcing incentive mechanisms to improve indoor geolocation accuracy, based on RSSI fingerprints and data protection parameters.
[0009] The invention is defined by the independent claims. The embodiments are defined by the dependent claims. The present invention improves the situation.
[0010] A first aspect of the invention relates to a method for comparing a first trajectory T with a second trajectory T', the trajectories T and T' being at least partly included in a geographical area, the geographical area being covered by a set of cells, the method being implemented by a computer and comprising: a determination of a first signature vector, Sig(EV), determined by applying a hash function, f, to an encrypted vector, EV, such that at least a k-th component of the first signature vector corresponds to the encrypted form of the minimum component of a vector resulting from the product of a vector V and a vector resulting from the application to a vector W= (W m ) m∈[1;n] of a k-th permutation included in a subset of h elements of the set of permutations of n values, where W m is equal to m and the encrypted vector, EV, being the result of the encryption of the vector V representing the trajectory T, the i-th component V i of the vector V = (V i ) i∈[1,n] being: of strictly positive real value α, when the trajectory T comprises an i-th transition of a set of n transitions, each transition of the set of transitions representing a transition from a cell of the geographical area to another cell of the geographic area;of strictly positive real value β with n.α< β, when the trajectory T does not include the i-th transition of the set of transitions; a comparison of Sig(EV) with a second signature vector Sig(EV'), the comparison comprising the application of a function D which, to two vectors EU and EU' associates an encrypted value of d(U,U'), where U and U' are two vectors and EU and EU' are respectively the encrypted vectors of U and U' and d is a distance, the second signature vector Sig(EV') corresponding to the second trajectory T'. ;
[0011] Thus, on the basis of encrypted data, namely the encrypted vector EV, obtained from data (namely the vector V) representing a first trajectory T, a comparison between the first trajectory and the second trajectory is carried out, more precisely a comparison between the data representing the first trajectory T and the data representing the second trajectory. The proposed method makes it possible to carry out such a comparison more efficiently, i.e. by consuming fewer computing resources while ensuring significant protection of the identity of the user at the origin of the first trajectory T.
[0012] Indeed, the hash function as defined in the method in combination with the particular structure of the vector V makes it possible to obtain a first signature vector, Sig(EV), whose components are independent of the trajectories which are not included in the trajectory T. In other words, the result of the hash function does not take into account the components of real value β, but only the components of real value α. The hash function thus filters the relevant components.Since the accuracy of the comparison depends on the number of components in the first signature vector (i.e., the number of distinct permutations that are used in the MinHash-like algorithm) and the relevance of these components, the large difference between the real-valued β components and the real-valued α components allows the comparison that must be made to compute the signature vector components to be greatly simplified, while maintaining the same accuracy. The complexity of the comparison is therefore reduced.
[0013] Thus, the operator of a service requiring the analysis of data relating to geographical trajectories can therefore carry out an efficient comparison without directly using unencrypted data representing the first trajectory. For example, a user device (telephone, smartphone, etc.) can encrypt the data representing the first trajectory before it is transmitted to the service operator to carry out the comparison. Thus, in the absence of the unencrypted data representing the first trajectory, the operator of the service using the encrypted data will not be able to identify the individual who produced the data representing the first trajectory.
[0014] A signature vector is understood to mean a vector whose components are the result of a hash function applied respectively to the components of another vector. The components of this other vector may or may not be encrypted.
[0015] The second trajectory may be a trajectory known to the computer and to the service operator operating the computer, i.e., a trajectory whose unencrypted data representing the trajectory T' are known to the service operator and to the computer. The data representing the second trajectory T' are, for example, a vector V' defined in the same way as V, namely, that the i-th component V' i of the vector V' = (V' i ) i∈[1,n] is: of real value α when the trajectory T' includes the i-th transition of the set of n transitions; of real value β when the trajectory T' does not include the i-th transition of the set of transitions.
[0016] Thus, the data representing the second trajectory is accessible to the service operator or is used by the computer before it is encrypted. A known trajectory may be a trajectory generated by the computer (or more generally by the service operator) or may result from the movement of a user whose trajectory is known to the service operator.
[0017] The computer then encrypts each component V' i of the vector V' to obtain the encrypted vector EV'.
[0018] The second trajectory may also not be known to the computer and the service operator: for example, the computer may receive encrypted data representing the second trajectory from a user device. This data corresponds to the encrypted value of the vector V', denoted EV'. Thus, in the absence of the data V' representing the second trajectory, the service operator using the encrypted data will not be able to identify the individual who produced the data representing the second trajectory T'.
[0019] The comparison can be made relative to only a portion of the trajectory or to the entire trajectory. For example, if the first trajectory and the second trajectory have the same starting and ending point, a comparison of all the trajectories can be made. Similarly, the comparison can be made on only a portion of one of the trajectories, for example when the two trajectories have a point in common.
[0020] The first trajectory T and the second trajectory T' may have the same starting points and / or the same arrival points and / or the same common point. By two trajectories having the same point (starting, arrival, or common), it is understood that the points, corresponding to the starting point, the arrival point or the point in common, of each of the two trajectories are included in the same cell, among a set of several cells covering at least part of a geographical area. For example, the set of cells may form a spatial partition of the geographical area. A fine mesh of the plurality of cells makes it possible to better characterize a trajectory. A cell corresponds to a geographical sub-area of the geographical area.
[0021] Several trajectories can also be compared with the second trajectory. In this case, the possibility of precisely identifying the individual by means of his trajectory is limited.
[0022] An encrypted vector is understood to mean a vector whose components are the result of the encryption of the components of another vector. Encryption is carried out using a key which may be a private key, in the case of symmetric encryption, or a public key, in the case of asymmetric encryption.
[0023] A signature vector corresponding to a trajectory is understood to be a vector resulting from the application of a hash function to a vector representing the trajectory. The vector representing the trajectory can be an encrypted vector (for example EV' and EV) or an unencrypted vector (for example V' and V).
[0024] By vector (encrypted or not) representing a trajectory, we mean a vector whose elements, whether decrypted or not, make it possible to trace more or less precisely the trajectory from which the vector originates. A vector representing a trajectory corresponds to the data representing the trajectory.
[0025] By comparison of two signature vectors is meant any operation making it possible to determine whether the trajectories from which the signature vectors are derived (i.e. the trajectories represented by vectors whose hashing results in the signature vectors) are similar or close.
[0026] By similar trajectories is meant two trajectories (at least for the portion of these trajectories between the points of which the comparison is made) passing through the same cells (for example, in the same order) of the set of cells, the cells of the set of cells covering a geographical area (for example, forming a partition of this geographical area) within which the two trajectories are located (at least for the portion of these trajectories concerned by the comparison). It is then equivalent to saying that the result of the function D applied to two signature vectors representing two similar trajectories is a cipher of 0.
[0027] By close trajectories, we mean two trajectories passing through common cells (for example, in the same order), the number of common cells compared to the number of cells crossed by one of the two trajectories being greater than a threshold. It is then equivalent to saying that the result of the function D applied to two signature vectors representing two close trajectories is a number with a value less than a threshold, noted η.
[0028] We can define a set of transitions, each transition of which represents a passage from one cell in the set of cells to another cell in this set. We then consider that a trajectory passing from one cell to another cell then includes the transition corresponding to this passage. Such a transition set can contain all the possible transitions from each of the cells in the set of cells to each of the other cells. It can also be a subset of this set. This other cell can be an adjacent cell. It is also possible that this other cell is not an adjacent cell. Indeed, transitions between adjacent cells may not be defined in the trajectory due to, for example, poor radio coverage.
[0029] Thus, two similar trajectories have the same transitions. Two close trajectories have a number of common transitions, relative to the total number of transitions of the close trajectories (for example, relative to the total number of transitions included in the trajectory with the most transitions among the two close trajectories) or absolutely, greater than a predetermined threshold.
[0030] In other words, when we define the function D which, to two vectors EU and EU' associates a numerical value of d(U,U'), where U and U' are two vectors and EU and EU' are respectively the numerical vectors of U and U' and d is a distance, the equality between two vectors EU and EU' (that is to say that the trajectories corresponding to the two vectors are similar) corresponds to a result of the calculation of D(EU,EU') which corresponds to the numerical value of 0. Moreover, two vectors EU and EU' are close (that is to say that the trajectories corresponding to the two vectors are close) when the result of the calculation of D(EU,EU') corresponds to the numerical value of a value lower than a threshold, noted η.
[0031] According to one embodiment, provision is made for obtaining the encrypted vector, EV, on user equipment and then transmitting the encrypted vector, EV, to the computer.
[0032] User equipment means any equipment (e.g., telephone, smartphone, or laptop) that can obtain data representing the user's trajectory (e.g., equipment with a GPS chip), encrypt this data, and then transmit it to the computer. More generally, user equipment can be replaced by several devices performing one or two of these three functions, for example, a connected object such as a watch that records the user's trajectory, the trajectory data can then be loaded onto a computer to be encrypted, and then transmitted to the computer. It is also possible for the user equipment to only perform the trajectory recording function and transmit the trajectory data (i.e., V) to a server that then produces an encrypted vector (EV) and transmits this encrypted vector to the computer.The server then separates itself from the operator's system to prevent the user from being identified. The transmission of trajectory data (encrypted or not) can be carried out via any communication network, for example the internet. It is also possible for the data representing the user's trajectory to be obtained centrally by an intermediary operator, for example directly by a network operator who can geolocate the user equipment. This intermediary operator can also encrypt this trajectory data and transmit it to the service operator, i.e. to the computer to carry out the comparison.
[0033] According to one embodiment α = 1 and β = n+1.
[0034] Thus, obtaining vectors representing the trajectories, that is to say generating the data representative of the trajectory, presents little complexity and thus requires little computing capacity.
[0035] According to one embodiment, the encryption is obtained using a public key pk and an encryption function, Encrypt( . ,pk), of a secure two-party computation (2PC) scheme.
[0036] The encryption function associates, by means of the public key, a real value with an encrypted value, that is to say a code which requires a private key to be decrypted.
[0037] All components of encrypted vectors and all encrypted values are obtained with the same encryption function. That is, the same encryption scheme is used.
[0038] A secure two-party computation (2PC) scheme is an encryption scheme that allows two parties to jointly compute the result of a function using their respective inputs, such that each party does not obtain any information about the other party's input. This can be a scheme that employs probabilistic encryption. Using such a scheme to encrypt trajectory data allows operations to be performed more efficiently directly on the encrypted data (codes) without knowing the actual value from which the encrypted data originates. Thus, it is possible to implement the following functions: the ScalarMult function, which associates to the value pair (Eu,u') a value equal to Encrypt(u.u',pk), where Eu is the encrypted value of the real value u (namely, Eu = Encrypt(u.pk)) and u' is an unencrypted real value; the Min function, which associates to the h-tuple {Eu i} i∈[1,n] a value Encrypt(min i∈[1,n] (ui ), pk), where Eu i is the encrypted value of the i-th value ui of the set {ui} i∈[1,n]; the ArgMin function, which associates to the h-tuple {Eu i} i∈[1,n] a value Encrypt(argmin i∈[1,n] (ui ), pk), where Eu i is the encrypted value of the i-th value ui of the set {ui} i∈[1,n]; the Equal function, which associates to the pair of vectors (EU,EU') a value Encrypt(1,pk) if U=U' and Encrypt(0,pk) if U≠U' where EU is the vector whose elements are the encrypted elements of U and EU' is the vector whose elements are the encrypted elements of U';the Add function, which associates a value Encrypt(u+u',pk) with the pair (Eu, Eu'), where Eu and Eu' are respectively the encrypted values of the real values u and u'; the Less function, which associates a value Encrypt(1,pk) with the pair of encrypted values (Eu, Eu'), of 1 if u ≥ u' and a value of Encrypt(0,pk) otherwise; and the Mult function, which associates a value Encrypt(u.u',pk) with the pair of encrypted values (Eu, Eu') and where Eu and Eu' are respectively the encrypted values of the values u and u'.
[0039] An encrypted value of A, Encrypt(A,pk), is one of the values that can be obtained by encrypting the value A with the Encrypt function using the public key pk. In the case of probabilistic encryption, several values can correspond to Encrypt(A,pk), so a value Encrypt(A,pk) is one of them. It is sometimes written Encrypt(A,pk ;r) where r is a random number.
[0040] According to one embodiment, the h components Sig(EV) j∈[1,h] resulting from the application of the hash function, f, to the encrypted vector, EV, correspond respectively to the encrypted form of the minimum component of a vector resulting from the product of a vector V and a vector resulting from the application to a vector W= (W m ) m∈[1;n] of a permutation included in the subset of h elements. That is, for each j∈[1,h], Sig(EV) j∈[1,h] is the encrypted value of min i∈[1,n] (V i .π j,i ), where EV i is the encrypted value of V i and π j,i is the value of the i-th component of the vector resulting from the application to the vector W= (1, 2,...,n) of the j-th permutation included in the subset of h elements. That is, the result Sig(EV) of the hash function f is defined by: Sig EV = f EV = Min i ∈ 1 n ScalarMult EV i , π j ,i j ∈ 1 h .
[0041] Thus, all components of Sig(EV) are obtained by applying the same operations with h different permutations. The hash function then allows a more efficient comparison of the trajectories while maintaining good precision.
[0042] According to one embodiment, the computer determines the signature vector Sig(EV'), this corresponding to the second trajectory T', the i-th component of the signature vector Sig(EV') being determined by calculating the encrypted value of min i∈[1,n] (V' i , π j,i ), where, π j,i is the value of an i-th component of the vector resulting from the application to the vector W of a j-th permutation included in the subset of h elements and (V' i ) i∈[1,n] are the components of a vector V' representative of the second trajectory T', the i-th component V' i of the vector V' = (V' i ) i∈[1,n] being equal to: the real value α when the second trajectory T' includes the i-th transition of the set of n transitions; or the real value β when the second trajectory T' does not include the i-th transition of the set of transitions.
[0043] Thus, the computer calculates for each i, the value of min i∈[1 , n] (V' i , π j,i ) from the vector V'. Each of these values is then encrypted to obtain the components of Sig(EV') and allow the comparison between Sig(EV) and Sig(EV'), that is to say to carry out the comparison of the trajectories T' and T in the encrypted domain. In this case the trajectory T' is known to the computer. The trajectory T' can thus be chosen in a relevant manner according to the knowledge of the geographical area. When T' is determined or chosen, we then speak of a reference trajectory. Thus, when several trajectories unknown to the service operator (such as the first trajectory) are compared to a reference trajectory, it is for example possible for the service operator to evaluate the number of individuals who take this trajectory. It is also possible to determine the number of individuals who take very close or similar trajectories.
[0044] The second signature vector, Sig(EV'), can also be obtained by encrypting the vector V' to obtain the encrypted vector EV' and applying the hash function f, we then find: Sig(EV')=f(EV')=[Min i∈[1,n] (ScalarMult(EV' i, π j,i ))] j∈[1,h] .
[0045] According to one embodiment, the computer determines the second signature vector Sig(EV'), determined by applying the hash function, f, to an encrypted vector, EV', such that at least one k-th component of the second signature vector corresponds to the encrypted form of the minimum component of a vector resulting from the product of a vector V' and the vector resulting from the application to the vector W of the k-th permutation included in the subset of h elements, the encrypted vector, EV', being the result of the encryption of a vector V' representing the second trajectory T', the i-th component V' i of the vector V' = (V' i ) i∈[1,n] being equal to: the real value α when the trajectory T' includes the i-th transition of the set of n transitions; or the real value β when the trajectory T' does not include the i-th transition of the set of transitions.
[0046] Thus, the computer applies the hash function f to the encrypted vector EV', in order to obtain the h components Sig(EV') j∈[1,h] of the second signature vector Sig(EV'). The vector EV' is the encrypted vector V' representing the trajectory T'. The computer can obtain the encrypted vector EV' in the same way as the vector EV (for example, following encryption and transmission of the vector EV' by a user equipment). Thus, the computer and the service operator only have encrypted data representing the trajectory T'. Thus, they cannot identify the individual from whom the data representing the trajectory T' come. The computer then compares Sig(EV) and Sig(EV'), that is, it compares the trajectories T and T' in the encrypted domain without knowing the trajectories T and T'.Thus, when several trajectories unknown to the service operator are compared with each other, it is for example possible for the service operator to assess possible traffic congestion in a given geographical area without any trajectory being known to the service operator. For this purpose, the service operator can decipher the result of the comparison but not the data relating to the compared trajectories. To ensure that the service operator cannot decipher the data relating to the trajectories, it is also possible to transmit to a third-party operator the encrypted result of the comparison or of a possible count, the third-party operator then being responsible for deciphering these results and transmitting these deciphered results to the service operator.Since the service operator does not have the private key to decrypt the trajectory data, it is thus guaranteed that it will not be able to obtain this data in an unencrypted manner but only the result of the comparison or a possible count. The third-party operator may possibly be a user of the service, so the user via the service will be able, for example, to know the number of users who take the same trajectory without it being necessary for another operator to manipulate unencrypted data relating to the trajectories.
[0047] The comparison operation can be performed with multiple trajectories, for example, comparing multiple trajectories that are not known to the service operator with each other or with one or more of a set of reference trajectories.
[0048] According to one embodiment, provision is made to add the first trajectory T to a group of trajectories corresponding to the second trajectory T' depending on the result of the comparison.
[0049] So, depending on the result of the comparison between Sig(EV) and Sig(EV'), for example if the two trajectories T and T' are considered similar or close, the first trajectory T is added into a group of trajectories corresponding to the second trajectory.
[0050] By a group of trajectories corresponding to the second trajectory is meant a set of trajectories similar or close to the second trajectory T'. The trajectories of the group of trajectories corresponding to the second trajectory may have been added to the group in the same way as the first trajectory T. The second trajectory T' is then the representative of the group of trajectories corresponding to the second trajectory.
[0051] According to one embodiment, it is planned to compare N trajectories (T l< ) l ∈[1,N] with the second trajectory T' by means of the comparison of each of the N determined signature vectors (Sig(EV l< )) l ∈[i,N] with the second signature vector Sig(EV'), the method further comprising a count of a number of trajectories included in the group of trajectories corresponding to the second trajectory T' among the N trajectories, the count comprising the calculation of a numerical value of the number, ecount[N], of trajectories included in the group of trajectories corresponding to the second trajectory T' among the N trajectories.
[0052] The determined signature vectors (Sig(EV l< )) l ∈[1,N] are obtained in the same way as the first signature vector Sig(EV), for example, by applying the hash function f to each of the N EV vectors l< . Each of the vectors obtained is then compared with Sig(EV'), that is to say that each of the N trajectories (T l< ) l ∈[1,N] is compared in the encrypted domain to the second trajectory T'. Depending on the encrypted result of the comparison between Sig(EV') and each of the Sig(EV l< ) l , for example if the trajectories T l< and T' are considered similar or close, the trajectory T l< is added to the group of trajectories corresponding to the second trajectory T'. Then, a count of the number of trajectories in the group of trajectories corresponding to the second trajectory T' is carried out in the encrypted domain. Thus, since the count is carried out without decrypting the result of the comparison, the number of trajectories in a group can therefore be calculated without the service operator being able to deduce it. The decryption of the result of the count, namely ecount[N], can then be done by another authorized operator (for example the third-party operator) or by the computer. For this, the anonymization of the users (those from whom the (T l< ) l ∈[1,N] trajectories) must be guaranteed, for example, if we can ensure that the number of trajectories in the counted group is necessarily large, thus preventing the identity of the users from being deduced.
[0053] When several groups corresponding to different trajectories are considered, each trajectory T l< is compared with each representative of each of these different groups. In this case a trajectory T l is added to a group if the trajectory T l< is equal to or closer to (depending on the distance d applied to the vector representing T l< and the vector representing the trajectory representing the group) of the group representative than of the other representatives.
[0054] According to one embodiment, the calculation of the numerical value of the number, ecount[N], of trajectories included in the group of trajectories corresponding to the second trajectory T' comprises the calculation of: Ecount k = Add ( ecount k − 1 , Equal Sig EV k , Sig EV ′ where, ecount[k] is the numerical value of a number of trajectories counted in the group of trajectories corresponding to the second trajectory T' among the trajectories (T l< ) l ∈[1,k] .
[0055] This ensures that the count is performed without decrypting the result of the comparison or the intermediate values of the count, namely ecount[k], which remain encrypted. The number of trajectories in a group cannot therefore be deduced by the service operator.
[0056] According to one embodiment, the counting of the number of trajectories comprises the application of a threshold function S to the numerical value ecount[N] of the number of trajectories among the N trajectories which are included in the group of trajectories corresponding to the second trajectory T', the function S associating with a numerical value Eu the numerical value Eu when u is greater than a threshold µ greater than or equal to 2 and the numerical value of a real value strictly less than µ otherwise, where Eu is the numerical value of the real value u.
[0057] Thus, the result of the threshold function is a numerical value that is not representative of the number of trajectories in a group corresponding to the second trajectory when this number is less than a threshold u. Thus, if the number of trajectories counted is less than the threshold µ, it is not possible to obtain the value of the number of trajectories in the group. Knowing that a group corresponding to a trajectory includes a very small number of trajectories can make it possible to identify the individual at the origin of the data relating to one of the trajectories in the group. Indeed, as previously mentioned, if the group includes, for example, a single trajectory close or even similar to the representative trajectory, then it is possible to determine significant places in the individual's life and possibly identify this individual.If the group includes many trajectories then it becomes impossible or at least more difficult to identify an individual via the representative trajectory of this group. Thus, the threshold function prevents us from being able to obtain the value, even numerically, of the number of trajectories in a group including too few trajectories and thus prevents us from being able to identify individuals in this group.
[0058] The threshold function can be defined by: S(Eu) = Mult(Less(Eu,µ), Eu).
[0059] According to one embodiment, provision is also made for decrypting the result S(Encount[N]) of the threshold function S applied to the encrypted value Ecount[N].
[0060] The decryption of the result S(Encount[N]) can be carried out by another operator (for example, the third-party operator), but also by the service operator because the threshold function S makes it possible to guarantee the anonymization of users (those from whom the (T l ) l ∈[1,N] trajectories) when µ is chosen sufficiently large.
[0061] According to another aspect, there is provided a computer program comprising instructions for implementing all or part of a method as defined herein when these instructions are executed by a processor. According to another aspect of the invention, there is provided a non-transitory, computer-readable recording medium on which such a program is recorded.
[0062] According to another aspect, there is provided a computer configured to compare a first trajectory T with a second trajectory, the trajectories T and T' being at least partly included in a geographical area, the geographical area being covered by a set of cells, the computer comprising: a processor; and a memory storing instructions, such that when these instructions are executed by the processor, they configure the user terminal to: determine a first signature vector, Sig(EV), by applying a hash function, f, to an encrypted vector, EV, such that at least one k-th component of the signature vector corresponds to the encrypted form of the minimum component of a vector resulting from the product of a vector V and a vector resulting from the application to the vector (1, 2,...,n) of a k-th permutation included in a subset of h elements of the set of permutations of n values, the encrypted vector, EV, being the result of the encryption of the vector V representing the trajectory T, the i-th component V i of the vector V = (V i ) i∈[i,n] being: of strictly positive real value α, when the trajectory T comprises an i-th transition of a set of n transitions, each transition of the set of transitions representing a transition from a cell of the geographical area to another cell of the geographical area; of strictly positive real value β with n.α< β, when the trajectory T does not include the i-th transition of the set of transitions; compare Sig(EV) with a second signature vector Sig(EV'), the comparison comprising the application of a function D which, to two vectors EU and EU' associates an encrypted value of d(U,U'), where U and U' are two vectors and EU and EU' are respectively the encrypted vectors of U and U' and d is a distance, the second signature vector Sig(EV') corresponding to the second trajectory T'. .
[0063] Other characteristics, details and advantages of the invention will appear on reading the detailed description below, and on analyzing the attached drawings, in which: there figure 1 illustrates the comparison of trajectories according to a particular embodiment; the figure 2 illustrates a flowchart representing the method according to a particular embodiment.
[0064] The example of the figure 1 describes a set of users 12, 22 and 32 having user terminals 11, 21 and 31. The user terminal 11 is described in more detail. The user terminal 11 comprises a communication network interface module (INT1) 14, a processor (PROC1) 15 and a memory (MEMO1) 16. The processor 15 is configured to determine the data (for example, the vector V 1< representing the trajectory T 1< ) relating to the geographical trajectory 10. The processor 15 is configured to encrypt this data to obtain the encrypted vector EV 1< . The processor 15 is further configured to control the communication network interface module (INT1) 14 in order to be able to transmit the encrypted vector EV 1< 13 to a computer 50 (for example, a server of the service operator) via the communication network 40, for example the Internet network.
[0065] The interface module with the communication network (INT1) 14 allows the user terminal 11 to transmit the encrypted vector EV 1 < 13 via the communication network 40. It may be an internet modem and / or a network card for a computer or even a wireless communication module making it possible to code and transmit a radio signal coming from an antenna of the communication network 40.
[0066] The memory 16 comprises a non-volatile memory on which the computer program is stored and a volatile memory on which the parameters for implementing the proposed method are stored, for example, the vector V 1< representing the trajectory T 1< , data relating to a division of a geographical area into cells, data relating to associations between pairs of cells and the corresponding transitions of a set of n transitions (ti ) i∈[1,n] and a public key.
[0067] The user terminals 11, 21 and 31 may be, for example, computers, smartphones or even digital tablets. The data relating to the geographical trajectories 10, 20 and 30 may be obtained by the user terminals 11, 21 and 31 by any techniques making it possible to geolocate a user terminal in a geographical area, for example, via a satellite geolocation system such as GPS or by a non-satellite geolocation system, for example by triangulation of signals from a radio telecommunications network. In this example, N encrypted vectors EV 1< to EV N< are transmitted to the computer 50, the users producing the encrypted vectors EV 3< to EV N-1< are not represented on the figure 1 .
[0068] The computer 50 comprises an interface module with the communication network (INT) 51, a processor (PROC) 52 and a memory (MEMO) 53.
[0069] The memory 53 comprises a non-volatile memory on which the computer program is stored and a volatile memory on which the parameters for implementing the invention are stored, for example, the encrypted vectors EV 1< to EV N< , the signature vectors of the reference trajectories, the result of a comparison between signature vectors, the functions f, D, Ecount and S, etc.
[0070] The communication network interface module (INT) 51 allows the computer 50 to receive the encrypted vectors EV 1< 13 to EV N< 33 via the communication network 40. It may be an internet modem and / or a network card for a computer or even a wireless communication module allowing a radio signal to be received and decoded from an antenna of the communication network 40.
[0071] The processor 52 is in particular configured to control the interface module with the communication network (INT) 51 in order to be able to receive the encrypted vectors EV 1 < 13 to EV N < 33 via the network 40 and possibly to send via the communication network 40 the result of a comparison of signature vectors or the result of a count Ecount(N) or even S(Ecount(N)). The processor 52 is also configured to determine the N signature vectors, (Sig(EV) l< ) l∈[1,N], from the N encrypted vectors EV 1< to EV N<. The processor 52 is configured to compare the N signature vectors, (Sig(EV) l< ) l ∈[i,N], between them or with other signature vectors relating to reference trajectories TR p∈[1,P]< . The processor 52 is configured to calculate Ecount(N) or S(Ecount(N)).
[0072] There figure 2 represents a flowchart of a method according to a particular embodiment.
[0073] In step S1, each user terminal 11, 21 and 31 respectively determines the trajectories T 1< , T 2< and TN< . For this, the processor 15 of the user terminal 11 can, based on GPS data and data relating to a division of a geographical area into cells, determine at each moment in which cell the user terminal 11 is located. On the basis of the data relating to associations between pairs of cells and the corresponding transitions, at each change of cell the processor 15 determines the transition corresponding to the change of cell. Each transition corresponds to the passage from one cell to another, that is to say that each transition corresponds to a pair of cells (for example (C 1 ,C 2 ) and to a direction of passage (for example from C 1 to C 2 ). Thus for each pair of cells (C 1 , C 2 ), two transitions are possible, namely the transition from C 1 to C 2 and from C 2 to C 1 .
[0074] In step S2, each user terminal 11, 21 and 31 respectively determines the vectors V 1< , V 2< and VN< representing each trajectory T 1< , T 2< and TN< . The processor 15 of the user terminal 11 determines the i-th component V 1< i of the vector V 1< as being equal to: to the real value α, for example “1” when the trajectory T 1< includes the i-th transition ti of the set of n transitions; to the real value β, for example “1+n” when the trajectory T 1< does not include the i-th transition ti of the set of transitions.
[0075] It is recalled that the real values α and β are strictly positive and are chosen such that n.α< β.
[0076] In step S3, each user terminal 11, 21 and 31 encrypts the components of the vectors V 1< , V 2< and VN< to obtain the encrypted vectors EV 1< , EV 2< and EV N< . Thus, the processor 15 of the user terminal 11 determines the i-th component EV 1< i of the vector EV 1< as being equal to Encrypt(V 1< i ,pk). The Encrypt function is an encryption function which can be that of a secure two-party computation (2PC) scheme and at least an encryption scheme making it possible to define the following functions: the ScalarMult function, which associates to the value pair (Eu,u') a value equal to Encrypt(u.u',pk), where Eu is the encrypted value of the real value u (namely, Eu = Encrypt(u.pk)) and u' is an unencrypted real value; the Min function, which associates to the h-tuple {Eu i} i∈[1,n] a value Encrypt(min i∈[1,n] (ui ), pk), where Eu i is the encrypted value of the i-th value ui of the set {ui} i∈[1,n]; the ArgMin function, which associates to the h-tuple {Eu i} i∈[1,n] a value Encrypt(argmin i∈[1,n] (ui ), pk), where Eu i is the encrypted value of the i-th value ui of the set {ui} i∈[1,n]; the Equal function, which associates to the pair of vectors (EU,EU') a value Encrypt(1,pk) if U=U' and Encrypt(0,pk) if U≠U' where EU is the vector whose elements are the encrypted elements of U and EU' is the vector whose elements are the encrypted elements of U';the Add function, which associates a value Encrypt(u+u',pk) with the pair (Eu, Eu'), where Eu and Eu' are respectively the encrypted values of the real values u and u'; the Less function, which associates a value Encrypt(1,pk) with the pair of encrypted values (Eu, Eu'), of 1 if u ≥ u' and a value of Encrypt(0,pk) otherwise; and the Mult function, which associates a value Encrypt(u.u',pk) with the pair of encrypted values (Eu, Eu') and where Eu and Eu' are respectively the encrypted values of the values u and u'.
[0077] In step S4, each user terminal 11, 21 and 31 sends the vectors EV 1< , EV 2< and EV N< respectively, via the internet network 40, to the computer 50 which receives them via its interface module with the communication network (INT) 51.
[0078] Steps S1 to S4 can also be performed by an intermediate operator, for example an operator of a communications network who can geolocate the different user terminals 11, 21 and 31, for example, by identifying in which cell of the operator's network each user terminal is located. The cells of the network operator's network can in this case correspond to the cells of the geographical area. Once the trajectories T 1< , T 2< and TN< are determined by the intermediate operator, the latter can determine the vectors V 1< , V 2< and VN< . Then the intermediate operator can transmit the encrypted vectors EV 1< , EV 2< and EV N< to the computer 50.
[0079] In step S5, the processor 52 of the computer 50 determines for each encrypted vector EV 1< , EV 2< and EV N< a signature vector Sig(EV 1< ), Sig(EV 2< ) and Sig(EV N< ). At least one j-th component of the Sig(EV') corresponds to the encrypted form of the minimum component of the vector resulting from the product of the vector EV l< and the vector resulting from the application to a vector W= (W m ) m∈[1;n] of the k-th permutation included in a subset of h elements. The j-th component Sig(EV l< ) j of the vector Sig(EV') being determined by f(EV') = Min i∈[1,n] (ScalarMult(EV l< i , π j,i )), where π j,i is the value of the i-th component of the vector resulting from the application to the vector (1, 2,...,n) of a j-th permutation included in a subset of h elements of the set of permutations of n values. Each vector Sig(EV') includes a number of components equal to the number of permutations included in the subset of h elements. Sig(EV') therefore includes h components. The number h is between 1 and n!. The higher h is, the more precise the comparison of the vector Sig(EV') with another signature vector will be, but the more the comparison requires computing resources; conversely, the lower h is, the less the comparison requires computing resources, but the less precise the comparison of the vector Sig(EV') with another signature vector will be.
[0080] In step S6, the processor 52 of the computer 50 compares each of the N signature vectors Sig(EV') with another signature vector.
[0081] The other signature vector may be a signature vector (also called a reference signature vector) determined on the basis of a vector representing a reference trajectory. For example, a trajectory generated by the computer or by the service operator according to geographical or strategic criteria using known techniques. More generally, each of the N signature vectors Sig(EV') is compared with a set of P reference signature vectors. These P reference signature vectors may be determined by applying the function f respectively to each of the encrypted vectors representing the reference trajectories (TR p< ) p∈[1,P] . Equivalently, these P signature vectors may be determined by encrypting the vector resulting from the application of the MinHash function to the unencrypted vectors representing the reference trajectories. The MinHash(U) being equal to (min(π 1 .U), min(π 2 .U),..., min(π h .U)), where π j is the vector resulting from the application to the vector (1, 2,...,n) of the j-th permutation included in the subset of h elements of the set of permutations of n values and U is an unencrypted vector. The MinHash function is further discussed in the paper "A Comparison of Hash-Based Methods for Trajectory Clustering" by M. Rayatidamavandi et al. published in the proceedings of the IEEE 2017 Int. Conf. 15th Intl Conf on Dependable, Autonomie and Secure Computing. . . In this case the comparison consists of determining the numerical value of the function D(Sig(EV l< ),Sig(EVR p< )), which corresponds to the numerical value of d(V l< ,VR p< ), where VR p< is the vector representing the reference trajectory TR p< .
[0082] It is also possible to compare the N signature vectors Sig(EV') with each other. The comparison also consists of determining the encrypted value of the function D(Sig(EV l< ),Sig(EV m< )), which corresponds to the numerical value of d(V l< ,V m< ), with the couple ( l ,m)∈[1,N] 2< and / #m.
[0083] In both cases, it is also possible to restrict oneself to determining whether two signature vectors correspond to similar trajectories or not, that is to say whether D(Sig(EV l< ),Sig(EVR p< )) or D(Sig(EV l< ),Sig(EV m< )), corresponds to the encrypted value Encrypt (0,pk) or not.
[0084] In step S7, the processor 52 of the computer 50 adds each of the N signature vectors Sig(EV') to one of the trajectory groups according to the result of the comparison.
[0085] When the N signature vectors Sig(EV') are compared with the reference signature vectors (Sig(EVR p< )) p∈[1,P], then the trajectory T' is added into the group of trajectories corresponding to the reference trajectory with which T' is closest, i.e. the reference trajectory TRP' such that the value D(Sig(EV l< ),Sig(EVR p'< )) corresponds to the numerical value of d(V l< ,VR p'< ) the lowest among the d(V l< ,VR p< ) p∈[1,P] . In other words, T' is added to the group corresponding to the reference trajectory TR p'< or p' = argmin(d(V l< ,VR p< ) p∈[1,P] ). That is to say, the numerical value of the function ArgMin i∈[1,p] is calculated (D(Sig(EV l< ),Sig(EVR p< )).
[0086] When the N signature vectors Sig(EV') are compared with each other, it is possible to create groups of trajectories corresponding to the trajectories having a dense neighborhood: that is, the most trajectories among the (T l< ) l ∈[1,N] in their surroundings, for example, at a distance d around them. That is, the groups to which the trajectories (T l< ) l ∈[1,N] , are the groups corresponding to trajectories (T' l< ) l ∈Σ (with T' l< = T' if l ∈ Σ and Σ subset of [1,N]) which are such that the cardinals of the groups {V l< / d(V l< ,V q< )<ρ, l ∈[1,N]}, with q∈Σ, are all greater than the cardinals of the groups {V l< / d(V l< ,V r< )<ρ, l ∈[1,N], r ∈ [1,N]\Σ}, with real ρ strictly greater than 0 and allowing to distinguish the different trajectories (T l< )l ∈[1,N] . The trajectories (T' l< ) l ∈Σ can be determined in the encrypted domain, i.e. without deciphering the vectors (EV l< ) l ∈[1,N] . Once the trajectories (T' l< ) l ∈Σ determined, the trajectories (T l< ) l ∈[1,N] are added to the groups corresponding respectively to the (T' l< ) l ∈Σ in the same way as in the case of reference trajectories TR p'< .
[0087] In both cases, it is also possible to add the trajectory T l< to the group for which the trajectory (whether it is reference or not) corresponding to the group is close to T l< , that is, D(Sig(EV l< ),Sig(EVR p< )) or D(Sig(EV l< ),Sig(EV m< )) corresponds to numerical values of values smaller than η.
[0088] In step S8, the processor 52 of the computer 50 counts in the encrypted domain the number of trajectories among the N trajectories (T l< ) l ∈[1,N] which are in each group corresponding to the reference trajectories or to the trajectories (T' l< ) l ∈Σ depending on the case.
[0089] In the case where we restrict ourselves to determining whether two signature vectors correspond to similar trajectories or not, that is to say if D(Sig(EV l< ),Sig(EVR p< )) or D(Sig(EV l< ),Sig(EV m< )), corresponds to the encrypted value Encrypt(0,pk) or not, then the count can be done by calculating: for each group, i.e. for each p∈[1,P], Ecount p< [N], by calculating by iteration Ecount[k]=Add(ecount[k-1], Equal(Sig(EV k< ), Sig(EVRP)); or for each group, i.e. for each l ∈Σ, Ecount'[N], by calculating by iteration Ecount[k]=Add(ecount[k-1], Equal(Sig(EV k< ),Sig(EV l< )).
[0090] In the case where the trajectories are compared with the reference trajectories, the calculations of step S8 can take into account the comparison step S6 and the addition step S7, thus, these steps are optional.
[0091] When we do not restrict ourselves to determining whether two signature vectors correspond to similar trajectories or not, for example, if we determine whether D(Sig(EV l< ),Sig(EVR p< )) or D(Sig(EV l< ),Sig(EV m< )) corresponds to numerical values of values smaller than η, then the count can be done by calculating: for each group, i.e. for each p∈[1,P], Ecount p< [N], by calculating by iteration Ecount p< [k]=Add[ecount p< [k-1];Equal(Less(Encrypt(η,pk),D(Sig(EV k< ),Sig(EVR p< )),Encrypt(1,pk))]; for each group, i.e. for each l ∈ Σ, Ecount'[N], by calculating Ecount l< [k]=Add[ecount l< [k-1];Equal(Less(Encrypt(η,pk),D(Sig(EV k< ),Sig(EV l< )),Encrypt(1,pk))] per iteration.
[0092] In step S9, the processor 52 of the computer 50 calculates for each group, the result of the function S applied to the result of the counting step S8, that is to say: for each p∈[1,P], the numerical value S(Encount p< [N]), or for each l ∈Σ, the encrypted value S(Encount l< [N]).
[0093] The function S associates with an encrypted value Eu the encrypted value Eu when u is greater than a threshold µ, for example 5, and a value Encrypt(0,pk) otherwise, where Eu is the encrypted value of the real value u.
[0094] This step may be optional, especially when the result of the count is not deciphered by the service operator and / or the computer. In addition, other techniques may be implemented to ensure the anonymization of users 12, 22 and 32.
[0095] In step S10, the result of the count is decrypted. When the decryption is carried out by the service operator, then step S9 can be carried out to guarantee the anonymization of users 12, 22 and 32. In this case the processor 52 of the computer 50 decrypts the result of the threshold function obtained in step S9, that is to say: for each p∈[1,P], the computer deciphers S(Encount p< [N]), or for each l ∈Σ, the computer deciphers S(Encount l< [N]).
[0096] When the decryption is not performed by the service operator, the computer transmits the encrypted values (Encount p< [N]) p ∈ [1,P] or S(Encount p< [N]) p ∈ [1,P] or the encrypted values (Encount l< [N]) l ∈Σ or S(Encount l< [N]) l∈Σ . These encrypted values are then decrypted by a third-party operator, for example, an operator authorized by users to process sensitive data, and in particular their data representing their trajectory 10, 20 and 30.
[0097] To decrypt these encrypted values, a private key is used, in accordance with asymmetric encryption techniques.
Claims
1. Method for comparing a first path T with a second path T', the paths T and T' being at least partly included in a geographical area, the geographical area being covered by a set of cells, the method being implemented by a computer and comprising: - determining a first signature vector, Sig(EV), determined by applying a hash function, f, to an encrypted vector, EV, so that at least a k-th component of the first signature vector corresponds to the cipher of the minimum component of a vector resulting from the multiplication of a vector V and of a vector resulting from the application to a vector W = (Wm)m∈[1;n] of a k-th permutation contained in a subset of h elements of the set of permutations of n values, where Wm is equal to m and the encrypted vector, EV, being the result of the encryption of the vector V representing the path T, the i-th component Vi of the vector V = (Vi)i∈[1,n] being: - of strictly positive real value α, when the path T comprises an i-th transition of a set of n transitions, each transition of the set of transitions representing a transition from one cell of the geographical area to another cell of the geographical area; - of strictly positive real value β with n.α < β, when the path T does not contain the i-th transition of the set of transitions; - comparing Sig(EV) with a second signature vector Sig(EV'), the comparison comprising applying a function D that, with two vectors EU' and EU", associates an encrypted value of d(U',U"), where U' and U" are two vectors and EU' and EU" are the encrypted vectors of U' and U", respectively, and d is a distance, the second signature vector Sig(EV') corresponding to the second path T'.
2. Method according to Claim 1, further comprising obtaining the encrypted vector, EV, on a user equipment and then transmitting the encrypted vector, EV, to the computer.
3. Method according to either of the preceding claims, wherein α = 1 and β = n+1.
4. Method according to one of the preceding claims, wherein the encryption is obtained by means of a public key pk and of an encryption function, Encrypt( . ,pk), of a secure two-party computation scheme, 2PC.
5. Method according to one of the preceding claims, wherein the hash function f is defined by: f EV = Min i ∈ 1 ,n ScalarMult EV i , π j ,i j ∈ 1 ,h ; where πj,i is the value of an i-th component of the vector resulting from the application, to the vector (1, 2,...,n), of a j-th permutation contained in a subset of h elements of the set of permutations of n values, where the function ScalarMult associates, with the pair of values (Ex,a), an encrypted value of a.x, where Ex is the encrypted value of the real value x, and the function Min associates, with the h-tuple of encrypted values {Exi}i∈[1,n], an encrypted value of mini∈[i,n](xi), where Exi is the encrypted value of xi.
6. Method according to one of the preceding claims, further comprising determining, by means of the computer, the signature vector Sig(EV') corresponding to the second path T', the i-th component of the signature vector Sig(EV') being determined by computing the encrypted value of mini∈[1,n](V'i, πj,i), where, πj,i is the value of an i-th component of the vector resulting from the application to the vector W of a j-th permutation contained in the subset of h elements and (V'i)i∈[1,n] are the components of a vector V' representative of the second path T', the i-th component V'i of the vector V' = (V'i)i∈[1,n] being equal to: - the real value α when the second path T' comprises the i-th transition of the set of n transitions; or - the real value β when the second path T' does not contain the i-th transition of the set of transitions.
7. Method according to one of Claims 1 to 5, further comprising determining, by means of the computer, the second signature vector Sig(EV'), determined by applying the hash function, f, to an encrypted vector, EV', so that at least a k-th component of the second signature vector corresponds to the cipher of the minimum component of a vector resulting from the multiplication of a vector V' and of the vector resulting from the application to the vector W of the k-th permutation contained in the subset of h elements, the encrypted vector, EV', being the result of the encryption of a vector V' representing the second path T', the i-th component V'i of the vector V' = (V'i)i∈[1,n] being equal to: - the real value α when the path T' comprises the i-th transition of the set of n transitions; or - the real value β when the path T' does not contain the i-th transition of the set of transitions.
8. Method according to one of the preceding claims, further comprising adding the first path T to a group of paths corresponding to the second path T' depending on the result of the comparison.
9. Method according to Claim 8, wherein N paths (Tl)l∈[1,N] are each compared with the second path T' by comparing each of N determined signature vectors (Sig (EVl))l∈[1,N] with the second signature vector Sig(EV'), the method further comprising counting a number of paths contained in the group of paths corresponding to the second path T' among the N paths, the counting comprising computing an encrypted value, ecount[N], of the number of paths contained in the group of paths corresponding to the second path T' among the N paths.
10. Method according to Claim 9, wherein computing the encrypted value, ecount[N], of the number of paths contained in the group of paths corresponding to the second path T' comprises computing: Ecount k = Add ( ecount k − 1 , Equal Sig EV k , Sig EV ′ ; where ecount[k] is the encrypted value of a number of paths counted in the group of paths corresponding to the second path T' among the paths (Tl)l∈[1,k], the function Add associates, with the pair of encrypted values (Ev', Ev"), an encrypted value of v'+v", where Ev' and Ev" are the encrypted values of the values v' and v", respectively, and the function Equal associates, with the pair of vectors (EU',EU"), an encrypted value of 1 if U' = U" and an encrypted value of 0 if U' ≠ U", where EU' is the vector the components of which are the encrypted elements of U', and EU" is the vector the components of which are the encrypted elements of U".
11. Method according to one of Claims 9 and 10, wherein counting the number of paths comprises applying a threshold function s to the encrypted value ecount[N] of the number of paths among the N paths that are contained in the group of paths corresponding to the second path T', the function s associating, with an encrypted value Eu, the encrypted value Eu when u is greater than a threshold µ greater than or equal to 2 and the encrypted value of a real value strictly less than µ, otherwise, where Eu is the encrypted value of the real value u.
12. Method according to Claim 11, wherein the threshold function s is the function defined by: S Eu = Mult Less Eu μ , Eu ; where the function Less associates, with the pair of encrypted values (Eu, Eu'), an encrypted value of 1 if u ≥ u' and of 0 otherwise, and the function Mult associates, with the pair of encrypted values (Eu, Eu'), an encrypted value of u.u', and where Eu and Eu' are the encrypted values of u and u', respectively.
13. Method according to either one of Claims 11 and 12, further comprising decrypting the result of the threshold function s applied to the encrypted value ecount[N].
14. Computer program comprising instructions for implementing the method according to any of the preceding claims, when these instructions are executed by a processor.
15. Computer configured to compare a first path T with a second path, the paths T and T' being at least partly included in a geographical area, the geographical area being covered by a set of cells, the computer comprising: a processor; and a memory storing instructions, so that when these instructions are executed by the processor, they configure the user terminal to: - determine a first signature vector, Sig(EV), by applying a hash function, f, to an encrypted vector, EV, so that at least a k-th component of the signature vector corresponds to the cipher of the minimum component of a vector resulting from the multiplication of a vector V and of a vector resulting from the application to the vector (1, 2,...,n) of a k-th permutation contained in a subset of h elements of the set of permutations of n values, the encrypted vector, EV, being the result of the encryption of the vector V representing the path T, the i-th component Vi of the vector V = (Vi)i∈[1,n] being: - of strictly positive real value α, when the path T comprises an i-th transition of a set of n transitions, each transition of the set of transitions representing a transition from one cell of the geographical area to another cell of the geographical area; - of strictly positive real value β with n.α < β, when the path T does not contain the i-th transition of the set of transitions; - compare Sig(EV) with a second signature vector Sig(EV'), the comparison comprising applying a function D that, with two vectors EU' and EU", associates an encrypted value of d(U',U"), where U' and U" are two vectors and EU' and EU" are the encrypted vectors of U' and U", respectively, and d is a distance, the second signature vector Sig(EV') corresponding to the second path T'.
