Opportunistic Networks (OppNets) can provide a low-cost and reliable way for the message forwarding in urban areas, especially in which traffic jams occur frequently. However, in terms of the OppNet based on the bicycle-sharing system (BSS), how to predict bicycle trips and improve routing performance still remains unsolved. Moreover, the exchange of auxiliary information among OppNet nodes (bike stations) will compromise the privacy of nodes/users. Thus we design the Two-Tier Probability Model (TTPM), including the InteR-day pattern and the IntrA-day pattern, to predict the trips accurately. Then the Discrete Optimization Differential Privacy (DODP) method is utilized to disturb the estimated InteR-day and IntrA-day probabilities, which will further protect the privacy of nodes and users. With TTPM and DODP, we propose an efficient privacy-preserving routing scheme for OppNet, which transforms the relay selection problem into the shortest path problem approximately. Extensive simulations show that the proposed routing scheme (TTPM) outperforms the benchmarks with the delivery ratio of more than 0.75 when the time-to-live is 5 days and the message generation rate is 6 pkts/hour. Compared with TTPM+Lap and TTPM+GRR, the proposed TTPM+DODP improves the delivery ratio by 30% and 3%, respectively.

While ride-hailing brings great convenience to our daily life, it also poses a threat to the passengers' location privacy. Although many perturbation-based methods have been proposed to protect the passengers' location privacy, imprecise can still result in poor assignments of the ride-hailing platform. In addition, the platform often has to balance multiple conflicting objectives when dispatching drivers. Thus, trading these objectives in an appropriate way is critical to the long-term development of the platform. In this paper, we focus on the assignment strategy design in ride-hailing, aiming to promote the performance of the dispatching system while protecting the passengers' location privacy. We first model the online order dispatching as a bipartite matching problem, where drivers are assumed to be offline available and orders arrive sequentially following a known distribution. Then, we construct several linear programs (LPs) to obtain the obfuscation matrix (for privacy-preserving) and the guiding solutions of assignments (for online dispatching). Finally, a parameterized LP-based online dispatching algorithm is proposed to flexibly trade the two assignment objectives, i.e., dispatch efficiency and fairness. Experimental results on real-world datasets demonstrate the effectiveness of our algorithms.