In this paper, we propose PP-OMDS (Privacy-Preserving OLAPbased Monitoring of Data Streams), an innovative framework for supporting the OLAP-based monitoring of data streams, which is relevant for a plethora of application scenarios (e.g., security, emergency management, and so forth), in a privacy-preserving manner. The paper describes motivations, principles and achievements of the PP-OMDS framework, along with technological advancements and innovations. We also incorporate a detailed comparative analysis with competitive frameworks, along with a trade-off analysis.
The PP-OMDS (Privacy-Preserving OLAP-based Monitoring of Data Streams)
framework focuses the attention on the research challenge represented by models, techniques
and algorithms for supporting privacy-preserving OLAP-based monitoring of data
streams. The investigated research context includes and integrates three distinct
components, namely privacy-preserving OLAP (e.g., [
Starting from this main motivation, the PP-OMDS framework aims at introducing models, techniques and algorithms for supporting privacy-preserving OLAP-based monitoring of data streams, which is relevant for modern distributed environments such as Clouds. This will fulfil actual limitations of state-of-the-art solutions that do not address the relevant application scenario represented by using OLAP tools and methodologies to support data stream monitoring in a privacy-preserving manner. Application scenarios of the PP-OMDS are many-fold (e.g., security and emergency management). Tangible results of the PP-OMDS framework in real-life applicative settings are represented by relevant research and innovation advancements in the context of models, techniques and algorithms for supporting privacy-preserving OLAP-based monitoring of data streams, which represent the main “result” of the framework. It should be noted that, nowadays, these topics play a critical role as they are “naturally” compliant with the EU H2020 research framework under the topic Big Data.
The methodology implemented within the PP-OMDS framework foresees a multistep approach that comprises: (i) conceptual analysis and design of case studies and related use cases showing in details how and according to which tasks users/applications interact with the PP-OMDS framework; (ii) conceptual analysis and design of models, techniques and algorithms for supporting privacy-preserving OLAP-based monitoring of data streams, which provides a top-down overview of the different PPOMDS framework’s components to be defined and (prototypically) implemented; (iii) design of the OLAP-based models and related algorithms for supporting the monitoring of data streams, which represents a more-detailed view of the activity (ii) specifically focused on OLAP-analysis aspects of the PP-OMDS framework; (iv) design of the privacy-preserving version of the OLAP-based models and related algorithms for supporting the monitoring of data streams, which represents a further refinement of models and algorithms defined by the activity (iii) but targeted to embedding privacy-preserving aspects of the PP-OMDS framework in such models and algorithms; (v) integration of the OLAP component and the privacy-preserving OLAP component, both oriented to data stream monitoring, into the PP-OMDS framework, according to several alternative computing models.
The scientific contribution of the research carried-out by the PP-OMDS framework is very high because state-of-the-art research, even though focused on the relevant problem of privacy-preserving OLAP over static data, lacks of proposals that are specifically focused on privacy-preserving OLAP over data streams, and, in addition to this, there is not a specific integration with monitoring aspects, which, contrary to this, are indeed very relevant for a wide spectrum of next generation data stream applications and systems.
Summarizing, the main research topics investigated by the PP-OMDS framework are the following: models, techniques and algorithms for supporting OLAP-based monitoring of data streams; models, techniques and algorithms for supporting the privacy-preserving version of OLAP-based monitoring of data streams; integration of models, techniques and algorithms devised in the context of the two previous topics/activities, according to several alternative computing models.
The paper describes motivations, principles and achievements of the PP-OMDS
framework, along with technological advancements and innovations. We also
incorporate a detailed comparative analysis with competitive frameworks, along with a
tradeoff analysis.
In the PP-OMDS framework, the main research focus is on the issue of supporting
privacy-preserving OLAP-based monitoring of data streams, which, has highlighted
above, is innovative in actual state-of-the-art research, and it is of relevant interest for
a wide spectrum of data stream applications (e.g., security [
With reference to the so-delineated application scenario, a critical challenge relies
on effectively and efficiently computing the OLAP view over sensor readings (i.e.,
streaming data), with the additional requirement that must be computed in a
privacypreserving manner (e.g., [
stream acquisition phase; (ii) My SQL database server provides the necessary data storage (even buffer-oriented) and management functionalities. Data Stream Monitoring Server – The server devoted to the monitoring of data streams, by implementing the algorithms proposed in the PP-OMDDS framework. It fully interacts with the Data Stream OLAP Server and the Data Stream Privacy-Preservation Server, respectively, in order to achieve the overall privacy-preserving OLAP-based monitoring of data streams pursued by the PP-OMDDS project. Technology used: Java programming language; JBoss application server. In particular, JBoss provides the necessary Java execution runtime environment for the Data Stream Monitoring Component, which is the component that implements classes and functions supporting the data stream monitoring phase.
Data Stream OLAP Server – The server devoted to support OLAP over of data streams, by implementing the algorithms proposed in the PP-OMDDS framework. It fully interacts with the Data Stream Monitoring Server and the Data Stream Privacy-Preservation Server, respectively, in order to achieve the overall privacy-preserving OLAP-based monitoring of data streams pursued by the PP-OMDDS project. Technology used: Java programming language; JBoss application server, Mondrian OLAP server. In particular: (i) JBoss provides the necessary Java execution runtime environment for the OLAPing Data Stream Component, which is the component that implements classes and functions supporting OLAP analysis over data streams; (ii) Mondrian OLAP server provides the necessary multidimensional data storage and management functionalities.
Data Stream Privacy-Preservation Server – The server devoted to support privacy-preserving management of data streams, by implementing the algorithms proposed in the PP-OMDDS project. It fully interacts with the Data Stream Monitoring Server and the Data Stream OLAP Server, respectively, in order to achieve the overall privacy-preserving OLAP-based monitoring of data streams pursued by the PP-OMDDS project. Technology used: Java programming language; JBoss application server. In particular, JBoss provides the necessary Java execution runtime environment for the Privacy-Preserving Data Stream Management Component, which is the component that implements classes and functions supporting the privacypreserving data stream management phase.
Analytics Server – The server devoted to support data stream analytics, according to the privacy-preserving OLAP-based vision pursued by the PPOMDS framework. Technology used: Java programming language; JBoss application server; MySQL database server. In particular: (i) JBoss provides the necessary Java execution runtime environment for the PP-OMDS Analytics Component, which is the component that implements classes and functions supporting the PP-OMDS analytics phase; (ii) My SQL database server provides the necessary data storage (even buffer-oriented) and management functionalities. Streaming Analytics Layer – The layer where streaming analytics functionalities are finally implemented and delivered to the client applications/users, based on specific analytics goals (e.g., emergency detection and management, security, etc.). Technology used: arbitrary (at the output layer).
The PP-OMDS framework exploits software and hardware solutions that are currently available and completely technologically-feasible (even considering open-source solutions, in the case of software). Indeed, looking at the hardware architecture of the PP-OMDS framework, we identify an (hardware) architecture that makes use of standard solutions for (i) collecting data streams, (ii) monitoring data streams, (iii) aggregating data streams, (iv) providing privacy-preserving methods over data streams, and (v) supporting OLAP-like query answering over data streams. All these components are commonly delivered on top of well-known architectures composed by (i) data servers, (ii) OLAP servers, (iii) application servers, (iv) client components (e.g., desktop computers, laptops, mobile devices, etc.). As regards the software architecture of the PPOMDS framework, we identify a (software) architecture populated by software components that can be developed by means of standard high-level programming languages (e.g., Java, C++, etc.) and standard data access and manipulation methods (e.g., JDBC, ODBC, etc.). It should be noted that, for both the hardware architecture and the software architecture of the PP-OMDS framework, the components to be developed adhere to well-assessed and mature technologies (both hardware and software) that clearly make the PP-OMDS framework completely-feasible. In addition to this, all the components of the PP-OMDS framework are already Cloud-enabled, hence the framework can be easily extended towards a Cloud-based application (hence, improving performance, reliability and availability). 4
The software product context of the PP-OMDS framework is represented by the wide area of streaming analytics tools and systems, and, in particular, those devoted to support stream monitoring. Nevertheless, it does not exist a solution that specifically focuses on the relevant problem of supporting privacy-preserving, OLAP-based monitoring of distributed data streams, as delineated by the PP-OMDS framework’s motivations. This confirms to us the innovativeness of our proposal.
In the following, we focus the attention on the specific block-founding problems that characterizes the PP-OMDS framework. For what regards the basic distributed data stream monitoring problem, some relevant tools and systems that are currently available are the following ones: APAMA Streaming Analytics, from Software AG; Stream Computing, from IBM; Event Stream Processor, from SAP; Apache Spark, from Cloudera.
For what regards both the basic OLAP analysis over data streams problem and the basic privacy-preserving data stream management problem, there do not exist direct tools and systems currently available, although vertical solutions on top of existing streaming analytics platforms can be devised.
The most critical limitation of competitor frameworks is represented by the fact that they all adhere to a common deployment model, i.e. providing a general platform for supporting data stream analytics. While on top of such (common) platform personalized solutions can still be developed (via ad-hoc IDE and functional programming environments), they do not focus on the specialized problem of supporting privacy-preserving OLAP-based monitoring of distributed data streams, despite the relevance of this problem and its target application scenarios (e.g., emergency management, security, and so forth).
On the other hand, developing personalized solutions on general-purpose streaming analytics platforms poses critical challenges for what regards a wide spectrum of issues, ranging from complexity overheads to (software) maintenance problems, from heterogeneous data format integration issues to analytical front-end tools that are not focused to the peculiarities of the target application (thus making the whole knowledge discovery from big streaming data harder), and so forth. All these considerations clearly suggest the adoption of an all-inside, self-contained analytical framework that, still being focused on the specific goal of supporting privacy-preserving OLAP-based distributed data stream monitoring, can be further specialized on particular, vertical application settings (still falling in the reference technological area).
In addition to this, the competitor frameworks do not provide an explicit support neither to OLAP analysis tools over data streams neither to privacy-preserving data stream management, making for them hard to follow the paradigms dictated by the PPOMDS framework’s paradigms. Table 1 proposes a summary on the comparative analysis of the competitor frameworks against the PP-OMDS framework, along the abovediscussed parameters/functionalities.
In this paper, we propose have proposed PP-OMDS, an innovative framework for supporting the OLAP-based monitoring of data streams, which is relevant for a plethora of application scenarios (e.g., security, emergency management, and so forth), in a privacy-preserving manner.