Localized differential privacy protection method based on label clustering

• Main Authors: ZHU YISHUI, DUAN ZONGTAO, HAO JIAHUAN, WANG QINGLONG, WANG LUYANG, FAN NA, CUI XUEYING
• Format: Patent
• Language: Chinese; English
• Published: 26.03.2021
• Subjects: ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDEDFOR ELSEWHERE; CALCULATING; CHECKING-DEVICES; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; GENERATING RANDOM NUMBERS; PHYSICS; REGISTERING OR INDICATING THE WORKING OF MACHINES; TIME OR ATTENDANCE REGISTERS; VOTING OR LOTTERY APPARATUS

The invention discloses a localized differential privacy protection method based on label clustering. Vehicle trajectory data mining and localized differential privacy protection are combined; beforeoriginal track data is sent to a server, localized differential privacy protection processing is carried out on the original track data. In the track division process, an improved window opening algorithm is used for dividing an original track, and track division is restrained by longitude and latitude, speed and other attributes; before label clustering, sub-track endpoints obtained after track division serve as nodes, a weighted undirected complete graph is constructed, and the clustering iteration result is more stable; in a random disturbance process, instantiated scene analysis is carriedout on privacy protection in vehicle track data mining, and a user responds to a problem whether a track point is an interest point or not through binomial distribution of [0, 1], that is, the problem is answered according to