Federated Learning (FL) is a novel client-server distributed learning framework that can protect data privacy. However, recent works show that FL is vulnerable to poisoning attacks. Many defenses with robust aggregators (AGRs) are proposed to mitigate the issue, but they are all broken by advanced attacks. Very recently, some renewed robust AGRs are designed, typically with novel clipping or/and filtering strategies, and they show promising defense performance against the advanced poisoning attacks. In this paper, we show that these novel robust AGRs are also vulnerable to carefully designed poisoning attacks. Specifically, we observe that breaking these robust AGRs reduces to bypassing the clipping or/and filtering of malicious clients, and propose an optimization-based attack framework to leverage this observation. Under the framework, we then design the customized attack against each robust AGR. Extensive experiments on multiple datasets and threat models verify our proposed optimizationbased attack can break the SOTA AGRs. We hence call for novel defenses against poisoning attacks to FL. Code is available at: https: //github.com/Yuxin104/BreakSTOAPoisoningDefenses.

Top-k frequent items detection is a fundamental task in data stream mining. Many promising solutions are proposed to improve memory efficiency while still maintaining high accuracy for detecting the Top-k items. Despite the memory efficiency concern, the users could suffer from privacy loss if participating in the task without proper protection, since their contributed local data streams may continually leak sensitive individual information. However, most existing works solely focus on addressing either the memory-efficiency problem or the privacy concerns but seldom jointly, which cannot achieve a satisfactory tradeoff between memory efficiency, privacy protection, and detection accuracy. In this paper, we present a novel framework HG-LDP to achieve accurate Top-k item detection at bounded memory expense, while providing rigorous local differential privacy (LDP) protection. Specifically, we identify two key challenges naturally arising in the task, which reveal that directly applying existing LDP techniques will lead to an inferior "accuracy-privacy-memory efficiency" tradeoff. Therefore, we instantiate three advanced schemes under the framework by designing novel LDP randomization methods, which address the hurdles caused by the large size of the item domain and by the limited space of the memory. We conduct comprehensive experiments on both synthetic and real-world datasets to show that the proposed advanced schemes achieve a superior "accuracy-privacy-memory efficiency" tradeoff, saving 2300× memory over baseline methods when the item domain size is 41,270. Our code is anonymously open-sourced via the link.

Property preserving encryption techniques have significantly advanced the utility of encrypted data in data outsourcing. However, while preserving certain properties (e.g., the prefixes or order of the data) in the encrypted data, such encryption schemes are typically limited to specific data types (e.g., IP addresses) or applications (e.g., range queries over order-preserved data), and highly vulnerable to the emerging inference attacks which may greatly limit their applications in practice. In this paper, to the best of our knowledge, we make the first attempt to generalize the prefix-preserving encryption to make it applicable to more general data types (e.g., geo-locations, market basket data, DNA sequences, numerical data and timestamps) and secure against the inference attacks. Furthermore, we present a generalized multi-view outsourcing framework that generates multiple indistinguishable data views in which one view fully preserves the utility for data analysis, and its accurate analysis result can be obliviously retrieved. We empirically evaluate the performance of our outsourcing framework against two common inference attacks on two different real datasets: the check-in location dataset and network traffic dataset. The experimental results demonstrate that our proposed framework preserves both privacy (with bounded leakage and indistinguishable data views) and utility (with 100% analysis accuracy).

In this paper, we propose a prefix-preserving encryption based data outsourcing framework which is applicable to multiple different types of data, such as geo-locations, market basket data, DNA sequences, numerical data and timestamps. It enables accurate data analyses on the encrypted data while ensuring strong privacy against inference attacks. The basic idea is to generates multiple indistinguishable data views in which one view fully preserves the utility for data analysis, and its accurate analysis result can be obliviously retrieved. We empirically evaluate the performance of our outsourcing framework against two common inference attacks on two different real datasets: the check-in location dataset and network traffic dataset, respectively. The experimental results demonstrate that our proposed framework preserves both privacy (with bounded leakage and indistinguishability of data views) and utility.

Topic mining extracts patterns and insights from text data (e.g., documents, emails and product reviews), which can be used in various applications such as intent detection. However, topic mining can result in severe privacy threats to the users who have contributed to the text corpus since they can be re-identified from the text data with certain background knowledge. To our best knowledge, we propose the first differentially private topic mining technique (namely TopicDP) which injects well-calibrated Gaussian noise into the matrix output of any topic mining algorithm to ensure differential privacy and good utility. Specifically, we smoothen the sensitivity for the Gaussian mechanism via sensitivity sampling, which addresses the major challenges resulted from the high sensitivity in topic mining for differential privacy. Furthermore, we theoretically prove the differential privacy guarantee under the Rényi differential privacy mechanism and the utility error bounds of TopicDP. Finally, we conduct extensive experiments on two real-word text datasets (Enron email and Amazon Reviews), and the experimental results demonstrate that TopicDP is a model-agnostic framework that can generate better privacy preserving performance for topic mining as compared against other differential privacy mechanisms.

Vertical Federated Learning (FL) is a new paradigm that enables users with non-overlapping attributes of the same data samples to jointly train a model without directly sharing the raw data. Nevertheless, recent works show that it's still not sufficient to prevent privacy leakage from the training process or the trained model. This paper focuses on studying the privacy-preserving tree boosting algorithms under the vertical FL. The existing solutions based on cryptography involve heavy computation and communication overhead and are vulnerable to inference attacks. Although the solution based on Local Differential Privacy (LDP) addresses the above problems, it leads to the low accuracy of the trained model. This paper explores to improve the accuracy of the widely deployed tree boosting algorithms satisfying differential privacy under vertical FL. Specifically, we introduce a framework called OpBoost. Three order-preserving desensitization algorithms satisfying a variant of LDP called distance-based LDP (dLDP) are designed to desensitize the training data. In particular, we optimize the dLDP definition and study efficient sampling distributions to further improve the accuracy and efficiency of the proposed algorithms. The proposed algorithms provide a trade-off between the privacy of pairs with large distance and the utility of desensitized values. Comprehensive evaluations show that OpBoost has a better performance on prediction accuracy of trained models compared with existing LDP approaches on reasonable settings. Our code is open source.

Log parsers first convert large-scale and unstructured system logs into structured data, and then cluster them into groups for anomaly detection and monitoring. However, the security vulnerabilities of the log parsers have not been unveiled yet. In this paper, to our best knowledge, we take the first step to propose a novel real-time black-box attack framework LogBug in which attackers slightly modify the logs to deviate the analysis result (i.e., evading the anomaly detection) without knowing the learning model and parameters of the log parser. We have empirically evaluated LogBug on five emerging log parsers using system logs collected from five different systems. The results demonstrate that LogBug can greatly reduce the accuracy of log parsers with minor perturbations in real time.

for all the predefined sensitive objects (e.g., humans and vehicles) in the video, and then propose a video sanitization technique VERRO that randomly generates utility-driven synthetic videos with indistinguishable objects. Therefore, all the objects can be well protected in the generated utility-driven synthetic videos which can be disclosed to any untrusted video recipient. We have conducted extensive experiments on three real videos captured for pedestrians on the streets. The experimental results demonstrate that the generated synthetic videos lie close to the original video for retaining good utility while ensuring rigorous privacy guarantee.

Privacy and security concerns often prevent the sharing of users' data or even of the knowledge gained from it, thus deterring valuable information from being utilized. Privacy-preserving knowledge discovery, if done correctly, can alleviate this problem. One of the most important and widely used data mining techniques is that of classification. We consider the model where a single provider has centralized access to a dataset and would like to release a classifier while protecting privacy to the best extent possible. Recently, the model of differential privacy has been developed which provides a strong privacy guarantee even if adversaries hold arbitrary prior knowledge. In this paper, we apply this rigorous privacy model to develop a Naive Bayes classifier, which is often used as a baseline and consistently provides reasonable classification performance. We experimentally evaluate the proposed approach, and discuss how it could be potentially deployed in PaaS clouds.

Web search logs contain extremely sensitive data, as evidenced by the recent AOL incident. However, storing and analyzing search logs can be very useful for many purposes (i.e. investigating human behavior). Thus, an important research question is how to privately sanitize search logs. Several search log anonymization techniques have been proposed with concrete privacy models. However, in all of these solutions, the output utility of the techniques is only evaluated rather than being maximized in any fashion. Indeed, for effective search log anonymization, it is desirable to derive the outputs with optimal utility while meeting the privacy standard. In this paper, we propose utility-maximizing sanitization based on the rigorous privacy standard of differential privacy, in the context of search logs. Specifically, we utilize optimization models to maximize the output utility of the sanitization for different applications, while ensuring that the production process satisfies differential privacy. An added benefit is that our novel randomization strategy maintains the schema integrity in the output search logs. A comprehensive evaluation on real search logs validates the approach and demonstrates its robustness and scalability.

Clustering of search engine queries has attracted significant attention in recent years. Many search engine applications such as query recommendation require query clustering as a pre-requisite to function properly. Indeed, clustering is necessary to unlock the true value of query logs. However, clustering search queries effectively is quite challenging, due to the high diversity and arbitrary input by users. Search queries are usually short and ambiguous in terms of user requirements. Many different queries may refer to a single concept, while a single query may cover many concepts. Existing prevalent clustering methods, such as K-Means or DBSCAN cannot assure good results in such a diverse environment. Agglomerative clustering gives good results but is computationally quite expensive. This paper presents a novel clustering approach based on a key insight -- search engine results might themselves be used to identify query similarity. We propose a novel similarity metric for diverse queries based on the ranked URL results returned by a search engine for queries. This is used to develop a very efficient and accurate algorithm for clustering queries. Our experimental results demonstrate more accurate clustering performance, better scalability and robustness of our approach against known baselines.

User search query logs have proven to be very useful, but have vast potential for misuse. Several incidents have shown that simple removal of identifiers is insufficient to protect the identity of users. Publishing such inadequately anonymized data can cause severe breach of privacy. While significant effort has been expended on coming up with anonymity models and techniques for microdata, there is little corresponding work for query log data. Query logs are different in several important aspects, such as the diversity of queries and the causes of privacy breach. This necessitates the need to design privacy models and techniques specific to this environment. This paper takes a first cut at tackling this challenge. Our main contribution is to define effective anonymization models for query log data along with proposing techniques to achieve such anonymization. We analyze the inherent utility and privacy tradeoff, and experimentally validate the performance of our techniques.

With the vast increase in collection and storage of data, the problem of data summarization is most critical for effective data management. Since much of this data is categorical in nature, it can be viewed in terms of a Boolean matrix. Boolean matrix decomposition (BMD) has been used to provide concise and interpretable representations of Boolean data sets. A Boolean matrix can be expressed as a product of two Boolean matrices, where the first matrix represents a set of meaningful concepts, and the second describes how the observed data can be expressed as combinations of those concepts. Typically, the combination is only in terms of the set union. In other words, a successful Boolean matrix decomposition gives a set of concepts and shows how every column of the input data can be expressed as a union of some subset of those concepts. However, this way of modeling only incompletely represents real data semantics. Essentially, it ignores a critical component -- the set difference operation: a column can be expressed as the combination of union of certain concepts as well as the exclusion of other concepts. This has two significant benefits. First, the total number of concepts required to describe the data may itself be reduced. Second, a more succinct summarization may be found for every column. In this paper, we propose the extended Boolean matrix decomposition (EBMD) problem, which aims to factor Boolean matrices using both the set union and set difference operations. We study several variants of the problem, show that they are NP-hard, and propose efficient heuristics to solve them. Extensive experimental results demonstrate the power of EBMD.