Local Differential Privacy for Data Security and Privacy in Key Value Pair Data

Abstract

The utilization of local differential privacy (LDP) for collecting data on homogeneous data has primarily been researched. In practical applications, a number of data categories, such as key-value pairs was valued concurrently. These categories include frequency of keys and mean values using each key. Given that key-value data has two sides and client may have several key-value pairs, it is difficult to attain a satisfactory utility-privacy tradeoff when collecting key-value data using LDP. Large and small datasets cannot be handled by current LDP techniques due to their lack of scalability. Small datasets lack the information needed to determine statistical parameters, while large datasets found in streaming data increase the possibility of data leakage because of the abundance of readily available information. newline