AI-Driven Adaptive Differential Privacy Framework for Secure Location-Based Services

Location-Based Services (LBS) is an integral component of the modern digital environment, enabling the delivery of applications like smart city management and personalized mobile services. However, the service is challenged by privacy concerns due to the perpetual requirement for accurate spatial-temporal data. Although conventional privacy protection mechanisms like Differential Privacy (DP) offer strong theoretical guarantees of privacy, the approach is often based on a static privacy budget (ε) that does not offer satisfactory results when considering the data utility-privacy trade-off across various applications. In the present study, an adaptive Differential Privacy approach is proposed, enabled by the power of Artificial Intelligence (AI) technology. The approach allows the system to adapt the static privacy budget by adjusting the value of ε based on the contextual sensitivity of the query. Using the Random Forest algorithm, the system is able to evaluate the risks of the query by considering various parameters like the nature of the location, time factor, and user behavior. The experimental results based on a synthetic dataset for 5,000 location queries show that the proposed method significantly enhances the privacy utility trade-off. In specific terms, the proposed method reduces the success rate of inference attack by 49.2%, while utility is preserved with an increase in MAE by 63.7% for a controlled scenario. This demonstrates the applicability of the proposed method for adaptive and scalable privacy protection using ML and DP for LBS environments.