Search

CN-121580443-B - Edge computing privacy protection method based on differential privacy

CN121580443BCN 121580443 BCN121580443 BCN 121580443BCN-121580443-B

Abstract

The invention discloses an edge computing privacy protection method based on differential privacy, and belongs to the technical field of privacy protection. The method realizes cooperative optimization of privacy and efficiency through a four-step core flow, and comprises the steps of firstly obtaining task data, computing capacity data, wireless channel condition data and privacy protection requirement data, secondly processing the data, establishing a position and use mode dual-privacy protection model, fusing a differential privacy technology and a simulated annealing algorithm, designing a probability density function and a privacy measurement function, analyzing a user task unloading frequency, regenerating an confusion unloading proportion, an upper and lower boundary of an optimal confusion interval and a task unloading scheme after virtual mapping, finally determining a task allocation proportion, executing unloading operation and dynamically adjusting a strategy. The method realizes multi-dimensional privacy full coverage, effectively reduces malicious inference risk, and maintains dynamic balance of privacy protection and task efficiency while adapting to the edge computing resource limited characteristic.

Inventors

  • ZHANG CONG

Assignees

  • 四川华鲲振宇智能科技有限责任公司

Dates

Publication Date
20260508
Application Date
20260126

Claims (8)

  1. 1. The edge computing privacy protection method based on differential privacy is characterized by comprising the following steps of: s1, acquiring related data, wherein the related data comprise task data generated by a user, computing capacity data of local equipment and an edge server, wireless channel condition data and user privacy protection requirement data; s2, processing the data acquired in the S1, establishing a position privacy protection model and a use mode privacy protection model, fusing a differential privacy technology and a simulated annealing algorithm, designing a corresponding probability density function and a privacy measurement function, and analyzing the unloading frequency of a user task; s3, generating a confusion unloading proportion, upper and lower boundaries of an optimal confusion zone and a task unloading scheme after virtual mapping according to the data processing result; S4, analyzing the generated result, determining the task allocation proportion of local calculation and edge server calculation, executing task unloading operation, dynamically adjusting the task allocation proportion and virtual mapping strategy, and realizing balance of privacy protection and task efficiency; The establishing a location privacy protection model in the step S2 comprises the following sub-steps: S2.1, designing a probability density function of a confusion unloading proportion based on a differential privacy technology and acquired task data, wherein the function is limited in a set interval, and the total probability integral is 1; s2.2, constructing a privacy measurement function by adopting a cross entropy correlation method based on a designed probability density function, and calculating the privacy leakage degree P, wherein the calculation formula is as follows: wherein Q (r) is probability distribution of actual unloading proportion without differential privacy protection, and Pdf (r|r) is probability distribution of unloading proportion after confusion under a differential privacy mechanism; s2.3, fusing a simulated annealing algorithm and the constructed privacy metric function, establishing an optimal confusion interval selection model, and dynamically adjusting the upper and lower boundaries of the confusion interval; The establishing the usage pattern privacy protection model in step S2 includes the following sub-steps: s2.4, designing a probability density function of virtual mapping based on the obtained user task unloading frequency data, wherein the function is limited in a set interval; s2.5, quantifying the privacy attribute of the task by utilizing a designed probability density function and by utilizing the ratio relation of task unloading probability, and identifying the task which has influence on user portrait construction; S2.6, based on quantized task privacy attributes, carrying out targeted analysis on user task unloading frequency, and screening tasks with unloading frequency fluctuation amplitude smaller than a preset threshold.
  2. 2. The method according to claim 1, characterized in that step S1 comprises the sub-steps of: s1.1, task data generated by a user is obtained, wherein the task data comprises a task total amount and a task type; S1.2, acquiring computing capability data of local equipment and an edge server for supporting task processing based on the acquired task data, wherein the computing capability data comprises CPU clock cycles executed by the local equipment every second and computing capability parameters of the edge server; S1.3, acquiring wireless channel condition data comprising channel bandwidth, channel gain and background noise related parameters by combining task transmission requirements; S1.4, aiming at privacy protection requirements in task processing, obtaining protection importance degree parameters of a user on position privacy and use mode privacy.
  3. 3. The method according to claim 1, characterized in that step S3 comprises the sub-steps of: S3.1, calculating to obtain a confusion unloading proportion according to a probability density function and a privacy metric function in the position privacy protection model; S3.2, based on the calculated confusion unloading proportion, performing iterative computation through an optimal confusion interval selection model, and outputting the upper and lower boundaries of the optimal confusion interval; S3.3, combining the upper and lower boundaries of the optimal confusion zone and a user task unloading frequency analysis result of the use mode privacy protection model, executing virtual mapping processing on the screened tasks, and generating a task list after virtual mapping and a corresponding task unloading scheme.
  4. 4. The method according to claim 1, characterized in that step S4 comprises the sub-steps of: S4.1, determining a task allocation ratio of local calculation and edge server calculation according to the generated confusion unloading ratio and the upper and lower boundaries of the optimal confusion interval; s4.2, executing unloading operation according to the determined task allocation proportion and the virtual mapped task list, and executing randomization processing on tasks with unloading frequency fluctuation amplitude smaller than a preset threshold value; s4.3, monitoring energy consumption and time delay data in a task processing process after unloading operation, and dynamically adjusting task allocation proportion and virtual mapping strategy based on the monitoring data.
  5. 5. A method according to claim 3, wherein step S2.3 comprises: Constructing a target optimization function based on the energy consumption factor, the time delay factor, the privacy protection factor and the corresponding total task processing energy consumption, the total task processing time delay and the privacy disclosure degree P; Setting an initial temperature, a termination temperature and a cooling strategy based on the constructed target optimization function, and generating new upper and lower bound solutions of the confusion zone in the solution space; calculating the difference value of the objective function between the upper and lower bound solutions of the newly generated confusion zone and the current solution; Judging whether to accept the new solution according to the calculated objective function difference and a preset criterion; If the termination condition is not met, the temperature is reduced according to a cooling strategy, the steps of generating a new solution, calculating a difference value and judging acceptance are repeated until the temperature is reduced to the termination temperature, and the upper and lower boundaries of the optimal confusion interval are output.
  6. 6. The method according to claim 1, wherein step S2.6 comprises: preprocessing original user task unloading data, removing invalid data and converting the invalid data into a set format; loading the preprocessed user task unloading data to a storage frame, reading and processing the data, and outputting a key value pair containing a task identifier, a user identifier and unloading frequency; And (3) summarizing and sorting the data based on the output key values, and screening out tasks with unloading frequency fluctuation amplitude smaller than a preset threshold value.
  7. 7. A method according to claim 3, wherein step S3.3 comprises: The user task unloads tasks with the frequency fluctuation amplitude larger than a preset threshold value and directly executes virtual mapping processing; the user task unloads tasks with the frequency fluctuation amplitude smaller than a preset threshold value, and carries out randomization response judgment based on a probability density function of virtual mapping to determine whether to execute virtual mapping processing; Integrating the virtual mapping processing results of tasks with larger frequency variation and tasks with lower frequency variation, and generating a virtual mapped task list and a corresponding task unloading scheme.
  8. 8. The method according to claim 4, wherein step S4.3 comprises: Continuously collecting task processing energy consumption data of the local equipment in the task processing process, and calculating the local calculation energy consumption based on inherent parameters of the local equipment, the task quantity processed locally and the CPU cycle number required by the local equipment for processing each bit of task; synchronously collecting task processing time delay data of the edge server; Based on the collected energy consumption data and time delay data, adjusting probability density function parameters and privacy measurement function parameters in the position privacy protection model and the use mode privacy protection model; And according to the adjusted model parameters, the confusion unloading proportion and the virtual mapping strategy of the subsequent tasks are optimized, and the balance of privacy protection and task efficiency is maintained.

Description

Edge computing privacy protection method based on differential privacy Technical Field The invention relates to the technical field of privacy protection, in particular to an edge computing privacy protection method based on differential privacy. Background The edge calculation is used as a distributed calculation paradigm close to the terminal equipment, and by virtue of the advantages of low delay and high bandwidth utilization rate, the edge calculation is widely applied to various terminal intensive scenes and becomes a key hub for connecting a cloud end with the terminal equipment. The technology sinks the core capacities of data processing, storage and the like from the cloud to the network edge node, effectively reduces the data transmission distance and network congestion, greatly improves the task response speed, and is particularly suitable for scenes with higher requirements on real-time performance. Meanwhile, the differential privacy technology is used as a privacy protection means with mathematical provability, and by introducing controllable noise into data, disclosure of individual privacy information is avoided on the premise of guaranteeing the usability of the data, and the differential privacy technology is one of core technologies in the field of data privacy protection. The simulated annealing algorithm shows remarkable advantages in scenes such as parameter adjustment, scheme optimization and the like by virtue of the global optimization capability, and can help the system to jump out of a local optimal solution to find a better balance state. At present, edge computing, differential privacy and optimization algorithms are all mature applied in the respective fields, and research and practice of related technologies are continuously advancing, so that a solid foundation is laid for multi-technology fusion application. With the continuous popularization of edge computing applications, the demands of cooperative optimization of user privacy protection and task processing efficiency are increasingly prominent, but the prior art system still has a plurality of problems to be solved urgently. Firstly, the privacy protection dimension is single, most of the existing schemes only focus on a certain class of position privacy or data privacy, cannot cover the multi-dimensional requirements of user use mode privacy and the like at the same time, and cannot comprehensively resist privacy inference attacks of malicious nodes. Secondly, the application of the differential privacy technology in the edge computing scene lacks targeted optimization, and the characteristic of limited edge node resources is not fully combined, so that the privacy budget is unreasonable to be distributed, and the imbalance condition that the privacy protection intensity is insufficient or the task processing efficiency is greatly reduced is easy to occur. Furthermore, the existing scheme generally lacks effective fusion with a global optimization algorithm, cannot dynamically adjust a privacy protection policy and a task offloading scheme, and is difficult to maintain the balance of privacy protection and task efficiency in a dynamically changing edge computing environment. Finally, the data acquisition of part of schemes is not comprehensive enough, and the system fails to collect key data such as task characteristics, equipment capacity, channel conditions, user privacy requirements and the like, so that the follow-up model construction and scheme generation lack sufficient data support, and the suitability and effectiveness of the whole technical scheme are affected. The existence of the problems severely restricts the large-scale application of the edge computing technology in privacy-sensitive scenes, and a technical scheme which can comprehensively cover the multi-dimensional privacy protection requirements, adapt to the edge computing resource characteristics and realize dynamic optimization is needed. Disclosure of Invention The invention aims to overcome the defects of the prior art and provides an edge computing privacy protection method based on differential privacy. The aim of the invention is realized by the following technical scheme: the edge computing privacy protection method based on differential privacy comprises the following steps: s1, acquiring related data, wherein the related data comprise task data generated by a user, computing capacity data of local equipment and an edge server, wireless channel condition data and user privacy protection requirement data; s2, processing the data acquired in the S1, establishing a position privacy protection model and a use mode privacy protection model, fusing a differential privacy technology and a simulated annealing algorithm, designing a corresponding probability density function and a privacy measurement function, and analyzing the unloading frequency of a user task; s3, generating a confusion unloading proportion, upper and lower boundaries of an optimal