The enhancements in computation technologies in the last decades enabled businesses to analyze the data that is collected through their systems which helps to improve their services. However, performing data analytics remains a challenging task for small- and medium-scale companies due to the lack of in-house experience and computational resources. Data Analytics-as-a-Service (DAaaS) paradigm provides such companies outsourced data analytics, where a company that is specialized in data analytics serves its knowledge and computational resources to the other companies, which need data analytics for their businesses. A major challenge in DAaaS is preserving the privacy of the outsourced data, which might contain sensitive customer or employee information or the intellectual property of the outsourcing company. Leakage of sensitive information has several consequences both for outsourcing and service provider companies as legal obligations, loss of reputation, and financial loss. Therefore, a well functioning outsourced analytics service should achieve several data protection measures such as confidentiality, integrity, and availability. In this thesis, we focus on the preservation of confidentiality in data analytics-as-a-service applications. We select three analytics applications that are becoming popular in outsourced data analytics, which are process analytics, machine learning, and marketing analytics. Despite there exist several other techniques that are commonly used in outsourced data analytics, we decide to focus on the algorithms of process analytics, machine learning, and marketing analytics since the privacy concerns in these analytics have not been investigated thoroughly. In confidential data analytics-as-a-service, our goal is to achieve confidentiality by protecting input/output privacy and maintaining the correctness and efficiency of analytics computations. To protect the privacy of data we use two secure computation techniques, which are homomorphic encryption and secure multiparty computation. To assure correctness, we propose several hybrid protocol designs that minimize the loss of accuracy in computations. For the efficiency of our protocols, we use several optimization techniques that reduce the computation and communication costs of private data analytics. Our protocols show promising results for confidential data analytics in the outsourced setting. @en

Online advertising is a multi-billion dollar industry, forming the primary source of income for many publishers offering free web content. Serving advertisements tailored to users’ interests greatly improves the effectiveness of advertisements, and is believed to be beneficial to publishers and users alike. The privacy of users, however, is threatened by the widespread collection of data that is required for behavioural advertising. In this paper, we present BAdASS, a novel privacy-preserving protocol for Online Behavioural Advertising that achieves significant performance improvements over the state-of-the-art without disclosing any information about user interests to any party. BAdASS ensures user privacy by processing data within the secret-shared domain, using the heavily fragmented shape of the online advertising landscape to its advantage and combining efficient secret-sharing techniques with a machine learning method commonly encountered in existing advertising systems. Our protocol serves advertisements within a fraction of a second, based on highly detailed user profiles and widely used machine learning methods.

@en

Smart grids promise a more reliable, efficient, economically viable, and environment-friendly electricity infrastructure for the future. State estimation in smart grids plays a pivotal role in system monitoring, reliable operation, automation, and grid stabilization. However, the power consumption data collected from the users during state estimation can be privacy-sensitive. Furthermore, the topology of the grid can be exploited by malicious entities during state estimation to launch attacks without getting detected. Motivated by the essence of a secure state estimation process, we consider a weighted-least-squares estimation carried out batch-wise at repeated intervals, where the resource-constrained clients utilize a malicious cloud for computation services. We propose a secure masking protocol based on data obfuscation that is computationally efficient and successfully verifiable in the presence of a malicious adversary. Simulation results show that the state estimates calculated from the original and obfuscated dataset are exactly the same while demonstrating a high level of obscurity between the original and the obfuscated dataset both in time and frequency domain.

@en

We investigate the trade-off between privacy and solution quality that occurs when a k-anonymized database is used as input to the bin-packing optimization problem. To investigate the impact of the chosen anonymization method on this trade-off, we consider two recoding methods for k-anonymity: full-domain generalization and partition-based single-dimensional recoding. To deal with the uncertainty created by anonymization in the bin-packing problem, we utilize stochastic programming and robust optimization methods. Our computational results show that the trade-off is strongly dependent on both the anonymization and optimization method. On the anonymization side, we see that using single dimensional recoding leads to significantly better solution quality than using full domain generalization. On the optimization side, we see that using stochastic programming, where we use the multiset of values in an equivalence class, considerably improves the solutions. While publishing these multisets makes the database more vulnerable to a table linkage attack, we argue that it is up to the data publisher to reason if such a loss of anonymization weighs up to the increase in optimization performance.@en

Multi-party access control has been proposed to enable collaborative decision making for the protection of co-owned resources. In particular, multi-party access control aims to reconcile conflicts arising from the evaluation of policies authored by different stakeholders for jointly-managed resources, thus determining whether access to those resources should be granted or not. While providing effective solutions for the protection of co-owned resources, existing approaches do not address the protection of policies themselves, whose disclosure can leak sensitive information about, e.g., the relationships of co-owners with other parties. In this paper, we propose a privacy-preserving multi-party access control mechanism, which preserves the confidentiality of user policies. In particular, we propose secure computation protocols for the evaluation of multi-party policies, based on two privacy-preserving techniques, namely homomorphic encryption and secure function evaluation. An experimental evaluation of our approach shows its practical feasibility in terms of both computation and communication costs.

@en

Privacy-preserving data aggregation is growing in popularity due to the increasing amount of online services depending on user data. This information is privacy-sensitive, warranting the need for protection during data-processing. A wide variety of approaches have been considered to achieve privacy during the processing. Examples include differential privacy, masking, cryptographic techniques (e.g. using homomorphic encryption which enables data processing under encryption). In recent works, several approaches employing the latter privacy-preserving technique has been proposed that is proven to be secure in terms of sensitive data protection. However, the research mainly focuses mostly on efficiency rather than on the selected network topology. In contrast to existing work, we consider a decentralized network, where data can be aggregated without the presence of a central authority, such as an aggregator. We propose two novel protocols based on homomorphic encryption and secret sharing, respectively. Our analyses confirm our claims regarding high efficiency, scalability, and security.

@en

The increasing demand for data mining in business intelligence has led to a significant growth in the adoption of data mining as a service paradigm which enables companies to outsource their data and mining tasks to a cloud service provider. Despite the popularity of the paradigm, the companies hesitate to enable the cloud providers' access to their data considering customer privacy and intellectual property. In this paper, we propose a privacy-preserving two-party protocol which aims to mine direct sequential patterns from outsourced protected data. We focus on direct sequential pattern mining since it is a widely used primitive in business process analysis. Considering the accuracy and confidentiality, we choose encryption over statistical methods for data protection and processing. To be able to process the encrypted data, we adopt a homomorphic encryption scheme, ElGamal cryptosystem. The novelty of our scheme is that it introduces an encryption switching method that enables us to use both multiplicative and additive homomorphism on ElGamal cryptosystem. The results of our analyses show that our protocol is more efficient than the state-of-the-art proposals in terms of computational cost with a similar communication cost.@en

Online advertising is a rapidly growing industry, forming the primary source of income for many publishers that offer free web content. The practice of serving advertisements based on individuals' interests greatly improves the expected effectiveness of advertisements, and is believed to be beneficial to publishers and users alike. However, the widespread data collection required for such behavioural advertising sparks concerns over user privacy. In this paper, we present AHEad, a privacy-preserving protocol for Online Behavioural Advertising that ensures user privacy by processing data in encrypted form. AHEad combines homomorphic encryption with a machine learning method commonly encountered in existing advertising systems. Advertisements are served based on detailed user profiles, while achieving performance linear in the size of user profiles. To the best of our knowledge, AHEad is the first protocol that preserves user privacy in behavioural advertising while allowing the use of detailed user profiles and machine learning methods.@en

Online advertising forms the primary source of income for many publishers offering free web content by serving advertisements tailored to users’ interests. The privacy of users, however, is threatened by the widespread collection of data that is required for behavioural advertising. In this paper, we present BAdASS, a novel privacy-preserving protocol for Online Behavioural Advertising that achieves significant performance improvements over the state-of-the-art without disclosing any information about user interests to any party. BAdASS ensures user privacy by combining efficient secret-sharing techniques with a machine learning method commonly encountered in existing systems. Our protocol serves advertisements within a fraction of a second, based on highly detailed user profiles and widely used machine learning methods.@en

Online Behavioural Advertising (OBA), the practice of showing advertisements based on a person’s web browsing behaviour, has become a vital component of the ad-supported web. The tracking of users’ browsing behaviour that is needed for OBA, however, raises privacy concerns. We give an overview of the OBA landscape, and describe which user information is collected, which techniques are used to perform the collection, and how user information is shared between companies. Moreover, we discuss the privacy concerns that are raised by current OBA practices. After identifying privacy concerns, we describe a range of existing techniques to protect user privacy in online advertising. These techniques are compared based on their feasibility in the current advertising ecosystem, including the potential utility they offer advertising companies and how well they can be integrated with current trends in online behavioural advertising. Finally, we identify open problems in the protection of user privacy in online advertising.@en

The growing complexity of software with respect to technological advances encourages model-based analysis of software systems for validation and verification. Process mining is one recently investigated technique for such analysis which enables the discovery of process models from event logs collected during software execution. However, the usage of logs in process mining can be harmful to the privacy of data owners. While for a software user the existence of sensitive information in logs can be a concern, for a software company, the intellectual property of their product and confidential company information within logs can pose a threat to company's privacy. In this paper, we propose a privacy-preserving protocol for the discovery of process models for software analysis that assures the privacy of users and companies. For this purpose, our proposal uses encrypted logs and processes them using cryptographic protocols in a two-party setting. Furthermore, our proposal applies data packing on the cryptographic protocols to optimize computations by reducing the number of repetitive operations. The experiments show that using data packing the performance of our protocol is promising for privacy-preserving software analysis. To the best of our knowledge, our protocol is the first of its kind for the software analysis which relies on processing of encrypted logs using process mining techniques.

@en

Validation in a big software system can be managed by analysis of its behaviour through occasionally collected event logs. Process mining is a technique to perform software validation by discovering process models from event logs or by checking the conformance of the logs to a process model. A well-known algorithm in process mining to discover process models is alpha algorithm. However, while utilising alpha algorithm is useful for software validation, the existence of some sensitive information in the log files may become a threat for the privacy of users. In this work, we propose a protocol for privacy-preserving alpha algorithm on encrypted data. Our protocol aims to generate process models for a software without leaking any information about its users. It achieves same computational complexity with the original algorithm despite the additional computation overhead.@en