In the rapidly evolving online display advertising market, conversion rate (CVR) prediction models are typically updated daily using datasets enriched with recent conversion logs. However, a significant challenge is the time gap, often spanning days or weeks, between ad clicks and conversions. This issue, known as delayed feedback, results in false negatives in training data, creating a dilemma between label accuracy and data freshness. Existing methods for mitigating delayed feedback are limited, due to strong underlying assumptions, insufficient use of recent data without observed conversions, or implicit control over false negatives. To address this, we propose a simple framework that redefines CVR prediction under delayed feedback as an unsupervised domain adaptation (UDA) problem. Our method learns from fresh data while minimizing the impact of inaccurate labels, by integrating existing click-through rate (CTR) or CVR models with UDA algorithms. A customized pretraining step is also incorporated to effectively utilize recent observed conversions. Comprehensive experiments on three datasets showcase the proposed method's superiority over state-of-the-art approaches and its potential to benefit from advancements in CTR modeling. The code is available at https://github.com/ThunderbornSakana/DelayAdapter.

Understanding the privacy relevant characteristics of data sets, such as reidentifiability and joinability, is crucial for data governance, yet can be difficult for large data sets. While computing the data characteristics by brute force is straightforward, the scale of systems and data collected by large organizations demands an efficient approach. We present KHyperLogLog (KHLL), an algorithm based on approximate counting techniques that can estimate the reidentifiability and joinability risks of very large databases using linear runtime and minimal memory. KHLL enables one to measure reidentifiability of data quantitatively, rather than based on expert judgement or manual reviews. Meanwhile, joinability analysis using KHLL helps ensure the separation of pseudonymous and identified data sets. We describe how organizations can use KHLL to improve protection of user privacy. The efficiency of KHLL allows one to schedule periodic analyses that detect any deviations from the expected risks over time as a regression test for privacy. We validate the performance and accuracy of KHLL through experiments using proprietary and publicly available data sets.

Graph data publishing under node-differential privacy (node-DP) is challenging due to the huge sensitivity of queries. However, since a node in graph data oftentimes represents a person, node-DP is necessary to achieve personal data protection. In this paper, we investigate the problem of publishing the degree distribution of a graph under node-DP by exploring the projection approach to reduce the sensitivity. We propose two approaches based on aggregation and cumulative histogram to publish the degree distribution. The experiments demonstrate that our approaches greatly reduce the error of approximating the true degree distribution and have significant improvement over existing works. We also present the introspective analysis for understanding the factors of publishing the degree distribution with node-DP.

In this paper, we present DPSense, an approach to publish statistical information from datasets under differential privacy via sensitivity control. More specifically, we consider the problem of publishing column counts for high-dimensional datasets, such as query logs or the Netflix dataset. The key challenge is that as the sensitivity is high, high-magnitude noises need to be added to satisfy differential privacy. We explore how to effectively performs sensitivity control, i.e., limiting the contribution of each tuple in the dataset. We introduce a novel low-sensitivity quality function that enables one to effectively choose a contribution limit while satisfying differential privacy. Based on DPSense, we further propose an extension to correct the under-estimation bias, which we call DPSense-S. Experimental results show that our proposed approaches advance the state of the art for publishing noisy column counts and for finding the columns with the highest counts. Finally, we give the analysis and discussion for the stability of DPSense and DPSense-S, which benefits from the high correlation between quality function and error, as well as other insights of DPSense, DPSense-S, and existing approaches.