As a typical deep learning technique, Graph Convolutional Networks (GCN) has been successfully applied to the recommendation systems. Aiming at the leakage risk of user privacy and the problem of fraudulent data in the recommendation systems, a Privacy Preserving Recommendation and Fraud Detection method based on Graph Convolution (PPRFD-GC) is proposed in the paper. The PPRFD-GC method adopts encoder/decoder framework to generate the synthesized graph of rating information which satisfies edge differential privacy, next applies graph-based matrix completion technique for rating prediction according to the synthesized graph. After calculating user's Mean Square Error (MSE) of rating prediction and generating dense representation of the user, then a fraud detection classifier based on AdaBoost is presented to identify possible fraudsters. Finally, the loss functions of both rating prediction module and fraud detection module are linearly combined as the overall loss function. The experimental analysis on two real datasets shows that the proposed method has good recommendation accuracy and anti-fraud attack characteristics on the basis of preserving users' link privacy.

With the widespread application of location-based services, how to effectively publish trajectories while preserving users' privacy has become a critical challenge. Existing privacy-preserving trajectory publishing methods often face issues such as missing in synthetic trajectories, low data utility, and privacy leakage of users' social relationships due to trajectory correlation. To address these issues, a differentially private trajectory publishing method based on feature extraction and trajectory correlation (DP-FETC) is proposed in this paper. The method is composed of two synergistic core algorithms. First, a trajectory synthesis algorithm (SynFE) employs adaptive grid discretization to extract key statistical features—including weighted location, origin-destination (OD), and length distributions—to generate high-fidelity synthetic trajectories that maintain high data utility. Second, to mitigate the risk of social relationship disclosure, an innovative correlated trajectory protection algorithm (PreOD) identifies highly correlated trajectories using a novel metric based on stay duration at points of interest. It then applies a targeted perturbation exclusively to the OD points of these high-risk trajectories. This strategy effectively obscures social links while minimizing the impact on overall data quality. Experimental results on two real-world datasets validate that the proposed method has good data utility while providing robust privacy guarantees.

With the rapid progress of artificial intelligence (AI) technology in medical scenarios, it becomes a trend for medical services to adopt various AI algorithms for auxiliary diagnosis and health care of patients. However, medical data is often sensitive and possibly owned by multiple participants without the willingness of data sharing. To solve this problem under the vertical partition scenario of medical data, a differentially private distributed collaborative XGBoost method named DP-DCXGBoost was proposed and applied for disease classification in the paper. Initially, a reputation-based participant selection algorithm was designed, which evaluated the contribution of participants to the global model and used it for reputation calculation to select proper participants. Then, in the collaborative training phase, the proposed method utilized the local vertical dataset of each participant to calculate feature buckets and splitting gains in order to collaboratively construct a differentially private global XGBoost classification model. Finally, the experimental analysis for two real disease datasets showed that the proposed method had good classification accuracy on the basis of preserving participants' data privacy.