CN105631360A - Private data aggregating method based on multidimensional decomposition in sensor network - Google Patents

Abstract

The invention discloses a private data aggregating method based on multidimensional decomposition in a sensor network. In the sensor network, an over-disturbance phenomenon is possibly caused by a general differential privacy method based on a Laplace noise mechanism due to higher global sensitiveness, so as to destroy the effectiveness of aggregated data. According to the method provided by the invention, the differential privacy protection is realized by decomposing a single data stream into an exponential-weighted multidimensional data stream and adding independent noise on every dimension of data stream according to the local sensitiveness and the privacy estimation of every dimension. In comparison with a general aggregation process under the Laplace noise mechanism, by using the method, the better data effectiveness is provided while the privacy of a user at a same extent is guaranteed.

Description

Privacy data aggregation method based on multi-dimensional decomposition in sensor network

The technical field is as follows:

the invention belongs to the field of privacy protection, and particularly relates to a privacy data aggregation method based on multidimensional decomposition in a sensor network.

Background art:

various sensors in many novel sensor networks (such as smart grids, participatory sensing systems and the like) collect time sequence data related to users or environments and transmit the time sequence data to a server, and the server can perform aggregation operation on the data and use the externally aggregated results for higher-level statistical analysis and data mining operation. On the one hand, the data aggregation operation reduces the redundancy and transmission amount of data, thereby reducing the energy consumption and the delay time of the data. On the other hand, the data aggregation operation only transmits the critical or destination-related data to the upper nodes or servers, and the possibility of leakage of fine-grained data is reduced to a certain extent. Although fine-grained information cannot be directly accessed, detailed monitoring results can still be deduced from the changed real-time data aggregation results, for example, the electricity consumption information of residents collected by smart meters in a smart grid can be used for deducing the on-off state of the device, and the continuous calorie data collected by a wearable sensor can reveal the physical condition of a user. At present, a great deal of valuable research is carried out on the privacy protection problem in the data aggregation process in the sensor network, wherein the protection method based on the differential privacy ensures that the aggregation results of two adjacent data sets are very similar, so that an attacker is difficult to deduce a single data record by manipulating the aggregation results, and good data utility is provided while the user privacy is ensured. The differential privacy method based on the laplacian noise mechanism may cause an over-perturbation phenomenon due to a large global sensitivity, thereby destroying the utility of the data.

The invention content is as follows:

the invention aims to overcome the defects of the prior art and provides a private data aggregation method based on multi-dimensional decomposition in a sensor network, which can improve the utility of aggregated data under the condition of ensuring the user privacy to the same degree as that of a common method.

In order to achieve the purpose, the invention is realized by adopting the following technical scheme:

the privacy data aggregation method based on multi-dimensional decomposition in a sensor network comprises the following steps of firstly adding noise generated by a multi-dimensional noise decomposition mechanism to monitoring data generated in each time slot, and then carrying out privacy data aggregation by using the generated data:

step1 initializes: given a dimension base b and a global sensitivity g, calculating a dimension degree d, and calculating a local sensitivity s on each dimension_iWhere i is 1,2, …, d is the dimension after decomposition;

step2 data acquisition and decomposition: directly acquiring monitoring data through bottom hardware equipment to obtain an original monitoring value X corresponding to each time slot t_tAnd decompose it into d-dimensional data

Step3 data perturbation: according to local sensitivity s in each dimension_iAnd a privacy budget for generating d noise components which are subject to a Laplace distribution and are independent of each otherAnd add it toForm a noise result

Step4 privacy Convergence: combining d-dimensional convergence results of last time slotAnd d-dimensional noise results of the current time slotObtaining d dimension convergence result of current time slot

Step5 data synthesis and release: denoising results of each dimension according to different dimension weightsComprehensively obtain and externally release a convergence result R of privacy protection_t。

The invention further improves the method that Step1 comprises the following specific operations: the dimensionality base number b is a natural number which is more than or equal to 2, the global sensitivity g is the maximum estimation value of monitoring data, and the expression of the dimensionality d is

Local sensitivity s in each dimension_iIs expressed as

Where expression (2) indicates that the local sensitivity in the lower dimension i 1, 2.., d-1 is b-1, while the sensitivity in the highest dimension i d is determined jointly by the dimension base b, the global sensitivity g, and the dimension number d.

The invention is further improved in that the original monitoring value X in Step2_tComprises the following steps:

$X_t=\underset{i=1}{\overset{d}{\Σ}}{x}_t^i{b}^{i-1}---\left(11\right)$

wherein,is the original monitoring value X_tThe ith-dimensional component of (c), b^i-1Is the dimension weight of the ith dimension.

The invention further improves the method that Step3 comprises the following specific operations: generating d noise components satisfying differential privacy and being independent of each otherRequire thatObeying the probability density function to a laplacian distribution of equation (4):

$Lap\left(\mathrm{\λ}\right)~\frac{1}{2\mathrm{\λ}}{e}^{-\frac{\left|n\right|}{\mathrm{\λ}}}---\left(12\right)$

where n is a noise variable and the parameter λ is determined by the local sensitivity s in the corresponding dimension_iDetermined in conjunction with the privacy budget, i.e.

Noisy results in various dimensionsIs expressed as

$y_t^i=x_t^i+n_t^i---\left(13\right)$

Wherein,is the original monitoring value X_tThe (d) th-dimensional component of (a),is the noise component in the corresponding dimension.

The invention is further improved in that Step4 obtains the aggregation result of privacy protection in each dimensionIs composed of

$r_t^i=r_{t-1}^i+y_t^i---\left(14\right)$

Wherein,is the convergence result of the ith dimension of the last time slot, and the convergence result of each dimension when the time slot t is 0Is 0.

A further improvement of the invention is that Step5 is specifically operated to have different dimensional weights w_iFor each dimension of noise resultsComprehensively obtain a convergence result R published externally_tIs composed of

$R_t=\underset{i=1}{\overset{d}{\Σ}}{r}_t^i{b}^{i-1}---\left(15\right)$

Wherein, b^i-1Dimension weight w representing the ith dimension_i；

The monitoring value Y after privacy protection is restored from the published convergence result_tIs composed of

Y_t＝R_t-R_t-1(16)。

Compared with the prior art, the invention is realized by adopting the following technical scheme:

according to the privacy data aggregation method based on multi-dimensional decomposition in the sensor network, disclosed by the invention, a single data stream is decomposed into multi-dimensional data streams with exponential weights, and independent noise is added to each dimension of data stream according to the local sensitivity and privacy budget of each dimension to realize differential privacy protection. Compared with the convergence process under a general Laplace noise mechanism, the method provides better data utility while ensuring the same degree of user privacy.

Description of the drawings:

FIG. 1 is a diagram of a multi-dimensional decomposition noise mechanism;

FIG. 2 is a schematic diagram of a privacy data aggregation process based on multi-dimensional decomposition;

FIG. 3 is a comparison graph of noise variance between a multi-dimensional decomposition noise mechanism and a common Laplace noise mechanism under different global sensitivities;

fig. 4 is a comparison graph of average relative errors of multi-dimensional decomposition data aggregation and common laplacian data aggregation under different privacy budgets.

The specific implementation mode is as follows:

the present invention is described in further detail below with reference to the attached drawing figures.

The invention relates to a privacy data aggregation method based on multi-dimensional decomposition in a sensor network, which comprises the following steps:

step1 initializes: given a dimension base b (generally 2, consistent with a computer binary representation mode) and a global sensitivity g (the maximum value of a sensor monitoring value can be taken), obtaining a dimension d as

Local sensitivity s in each dimension_i(wherein i is 1,2, …, d) is

Step2 data acquisition and decomposition: referring to fig. 2, the monitoring data is directly obtained through the bottom hardware device, and the original monitoring value X corresponding to each time slot t is obtained_tAnd decomposing it into d-dimensional data, therebyCan be expressed as

$X_t=\underset{i=1}{\overset{d}{\Σ}}{x}_t^i{b}^{i-1}---\left(3\right)$

Wherein,is the original monitoring value X_tThe ith-dimensional component of (c), b^i-1Is the dimension weight of the ith dimension.

Step3 data perturbation: referring to FIG. 1, according to the local sensitivity s in each dimension_iAnd a privacy budget for generating d noise components satisfying differential privacy and independent of each otherObeying the laplace distribution with the probability density function as expression (4):

$Lap\left(\mathrm{\λ}\right)~\frac{1}{2\mathrm{\λ}}{e}^{-\frac{\left|n\right|}{\mathrm{\λ}}}---\left(4\right)$

where n is a noise variable and the parameter λ is determined by the local sensitivity s in the corresponding dimension_iDetermined in conjunction with the privacy budget, i.e.In practical application, the Laplace distribution can be replaced by random distribution functions such as geometric distribution and Gaussian distribution, and expected effects can be obtained;

noisy results in various dimensionsIs expressed as

$y_t^i=x_t^i+n_t^i---\left(5\right)$

Wherein,is the original monitoring value X_tThe (d) th-dimensional component of (a),is the noise component in the corresponding dimension;

step4 privacy Convergence: when the time slot t is 0, the result is converged by each dimensionIs 0; when time slot t>At 0, according to the convergence result of the ith dimension of the last time slotAnd the noise result of the ith dimension of the current time slotObtaining a convergence result of privacy protection of each dimensionIs composed of

$r_t^i=r_{t-1}^i+y_t^i---\left(6\right)$

Step5 data synthesis and release: referring to fig. 2, after privacy calculation at Step4, the noise results of each dimension are weighted according to different dimensionsComprehensively obtain a convergence result R published externally_t(the convergence target may be a sum or a count) of

$R_t=\underset{i=1}{\overset{d}{\Σ}}{r}_t^i{b}^{i-1}---\left(7\right)$

Wherein, b^i-1Dimension weight w representing the ith dimension_i；

The monitoring value Y after privacy protection is restored from the published convergence result_tIs composed of

Y_t＝R_t-R_t-1(8)

Referring to fig. 3, given that the global sensitivity range g is an integer between 2 and 2000, the privacy budget is 2, and the randomly selected dimension base b is 3, 8, and 1000, respectively, the variance of noise generated by the multidimensional decomposition noise mechanism used in the present invention is generally smaller than the variance of noise generated by the general laplacian noise mechanism, and a smaller noise variance may bring better data utility, which indicates that the multidimensional decomposition noise mechanism in the present invention has superiority in utility compared with the existing mechanism.

Referring to fig. 4, given the same global sensitivity g of 4083, the average relative error generated by the aggregation of multidimensional decomposition data and the aggregation of normal laplace data gradually decreases as the privacy budget increases. This reflects the nature of differential privacy, with a trade-off between privacy and utility, i.e., the smaller the privacy budget, the better the data privacy, and the lower the data utility, and vice versa. The randomly selected dimension cardinality b is 2 and 2042, the same privacy budget is given, the average relative error generated by the multi-dimensional decomposition data aggregation is lower than that of the common Laplace data aggregation, and the fact that the multi-dimensional decomposition data aggregation can bring better effectiveness is reflected.

In practical application, the performance of the convergence process can be improved by utilizing the characteristic integration cascade buffer counting method of multidimensional decomposition.

Claims (6)

1. The privacy data aggregation method based on multi-dimensional decomposition in the sensor network is characterized in that noise generated by a multi-dimensional noise decomposition mechanism is added to monitoring data generated in each time slot, and then the generated data is utilized to carry out privacy data aggregation, and the method specifically comprises the following steps:

step1 initializes: given a dimension base b and a global sensitivity g, calculating a dimension degree d, and calculating a local sensitivity s on each dimension_iWhere i is 1,2, …, d is the dimension after decomposition;

step2 data acquisition and decomposition: through the bottom layer hardDirectly acquiring monitoring data by piece equipment to obtain an original monitoring value X corresponding to each time slot t_tAnd decompose it into d-dimensional data

Step3 data perturbation: according to local sensitivity s in each dimension_iAnd a privacy budget for generating d noise components which are subject to a Laplace distribution and are independent of each otherAnd add it toForm a noise result

Step4 privacy Convergence: combining d-dimensional convergence results of last time slotAnd d-dimensional noise results of the current time slotObtaining d dimension convergence result of current time slot

Step5 data synthesis and release: denoising results of each dimension according to different dimension weightsComprehensively obtain and externally release a convergence result R of privacy protection_t。

2. The private data aggregation method based on multidimensional decomposition in the sensor network as claimed in claim 1, wherein Step1 is specifically operated as follows: the dimensionality base number b is a natural number which is more than or equal to 2, the global sensitivity g is the maximum estimation value of monitoring data, and the expression of the dimensionality d is

Local sensitivity s in each dimension_iIs expressed as

Where expression (2) indicates that the local sensitivity in the lower dimension i 1, 2.., d-1 is b-1, while the sensitivity in the highest dimension i d is determined jointly by the dimension base b, the global sensitivity g, and the dimension number d.

3. The private data gathering method based on multidimensional decomposition in sensor network as claimed in claim 1, wherein the original monitoring value X in Step2_tComprises the following steps:

$X_t=\underset{i=1}{\overset{d}{\Σ}}{x}_t^i{b}^{i-1}---\left(3\right)$

wherein,is the original monitoring value X_tThe ith-dimensional component of (c), b^i-1Is the dimension weight of the ith dimension.

4. The private data aggregation method based on multidimensional decomposition in the sensor network as claimed in claim 1, wherein Step3 is specifically operated as follows: generating d noise components satisfying differential privacy and being independent of each otherRequire thatObeying the probability density function to a laplacian distribution of equation (4):

$Lap\left(\mathrm{\λ}\right)~\frac{1}{2\mathrm{\λ}}{e}^{-\frac{\left|n\right|}{\mathrm{\λ}}}---\left(4\right)$

where n is a noise variable and the parameter λ is determined by the local sensitivity s in the corresponding dimension_iDetermined in conjunction with the privacy budget, i.e. $\mathrm{\λ}=\frac{s_i}{\mathrm{\ϵ}};$

Noisy results in various dimensionsIs expressed as

$y_t^i=x_t^i+n_t^i---\left(5\right)$

Wherein,is the original monitoring value X_tThe (d) th-dimensional component of (a),is the noise component in the corresponding dimension.

5. The method for gathering private data based on multidimensional decomposition in sensor network as claimed in claim 1, wherein Step4 obtains the gathering result of privacy protection in each dimensionIs composed of

$r_t^i=r_{t-1}^i+y_t^i---\left(6\right)$

Wherein,is the convergence result of the ith dimension of the last time slot, and the convergence result of each dimension when the time slot t is 0Is 0.

6. The method for gathering private data based on multidimensional decomposition in sensor network as claimed in claim 1, wherein Step5 is specifically operated to weight w according to different dimensions_iFor each dimension of noise resultsComprehensively obtain a convergence result R published externally_tIs composed of

$R_t=\underset{i=1}{\overset{d}{\Σ}}{r}_t^i{b}^{i-1}---\left(7\right)$

Wherein, b^i-1Dimension weight w representing the ith dimension_i；

The monitoring value Y after privacy protection is restored from the published convergence result_tIs composed of

Y_t＝R_t-R_t-1(8)。