Trajectory Data Publishing Method with Local Differential Privacy

The local differential privacy algorithm has strong privacy protection ability, which can effectively prevent privacy attacks by untrusted third parties and provide comprehensive protection for sensitive information. However, it also has some disadvantages such as the huge demand for data quantity and the low accuracy of published data. We consider that the attacker has all the background knowledge, and propose a permanent random response to the finite number of digits to solve the accuracy problem. At the same time, we also propose an instant random response to specific disturbance bits to resist vertical attacks and violent decryption. In terms of parameter adjustment, our work adopts the method of information theory and redefines the pri-vacy protection degree with mutual information by constructing the function of privacy protection degree and data availability. The utility is improved under the precondition of the expected privacy protection degree. And the average error of the data is controlled to about 2 meters, which improves the availability of the data by about 5 % compared to a method with the same level of protection. The utility is improved by about 25 % compared to the conventional randomized response.