Previous works about privacy preserving serial data publishing on dynamic databases have relied on unrealistic assumptions of the nature of dynamic databases. In many applications, some sensitive values changes freely while others never change. For example, in medical applications, the disease attribute changes with time when patients recover from one disease and develop another disease. However, patients do not recover from some diseases such as HIV. We call such diseases permanent sensitive values. To the best of our knowledge, none of the existing solutions handle these realistic issues. We propose a novel anonymization approach called HD-composition to solve the above problems. Extensive experiments with real data confirm our theoretical results.

Graph Neural Networks (GNNs) have been shown to be powerful in a wide range of graph-related tasks. While there exists various GNN models, a critical common ingredient is neighborhood aggregation, where the embedding of each node is updated by referring to the embedding of its neighbors. This paper aims to provide a better understanding of this mechanisms by asking the following question: Is neighborhood aggregation always necessary and beneficial? In short, the answer is no. We carve out two conditions under which neighborhood aggregation is not helpful: (1) when a node's neighbors are highly dissimilar and (2) when a node's embedding is already similar with that of its neighbors. We propose novel metrics that quantitatively measure these two circumstances and integrate them into an Adaptive-layer module. Our experiments show that allowing for node-specific aggregation degrees have significant advantage over current GNNs.

Accurate, quick, and reliable event detection is essential to both conventional and automated drilling processes. Methods for real-time detection of events using Bayesian inference have been described previously. These systems were initially designed to provide alarms to rig crews with little or no operator setup or maintenance. These requirements have resulted in certain limitations that restrict their use to relatively simple situations. For example, the drillstring washout detection method was based primarily on the behavior of hydraulic coefficients of a simple hydraulic model computed from standpipe pressure and flow, which meant that it was limited to only scenarios of drilling without a mud motor. New sensors and enhancements to the Bayesian inference techniques significantly improve event detection. Although applicable to a broader set of events, improvements to these detection methods can be explained using the example of a new drillstring washout detection module. The new module combines a method of detection of on-bottom drilling with enhanced signal characterization of standpipe pressure and flow to create an algorithm that can effectively handle drilling with a motor. Another improvement in the washout detection algorithm is the use of MWD turbine RPM when it is available. Unaffected by the differential pressure and depth changes, the turbine RPM data have proven to be more effective in detecting washouts occurring in the drillstring above the MWD tool. Testing both on data from several previously recorded washouts in drillstrings and on real-time data from wells with washout demonstrates the new capability of the algorithm to detect these difficult-to-characterize events while keeping false detections to a minimum.