PP-OMDS: An Effective and Efficient Framework for Supporting

Privacy-Preserving OLAP-based Monitoring of Data Streams

Alfredo Cuzzocrea

, Assaf Schuster

and Gianni Vercelli

University of Trieste and ICAR-CNR, Trieste, Italy

Technion, Haifa, Israel

University of Genoa, Genoa, Italy

Keywords: Privacy-Preserving OLAP over Data Streams, OLAP-based Monitoring of Data Streams, Privacy-Preserving

OLAP-based Monitoring of Data Streams.

Abstract: In this paper, we propose PP-OMDS (Privacy-Preserving OLAP-based 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.

1 INTRODUCTION

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., (Pernul et al., 2000;

Wang et al., 2004a; Wang et al., 2004b; Agrawal et

al., 2005; Hua et al., 2005; S.Y. Sung et al., 2006)),

data stream monitoring (e.g., (Bulut et al., 2009; Gan

& Dai, 2014; K.-Y. Cao et al., 2014; Huang et al.,

2006; Huang et al., 2007)), and privacy-preserving

data stream monitoring (e.g., (Dwork et al., 2010b;

T.-H.H. Chan et al., 2010; D.J. Mir et al., 2011;

Dwork et al., 2006; Beimel et al., 2008; Chen et al.,

2012; Friedman et al., 2014)). Even if some sporadic

works on supporting aggregate monitoring queries in

a privacy-preserving manner exist (e.g., (Shi et al.,

2011; Fan & Xiong, 2012)), the three related issues

have not been investigated together under the

umbrella of a common framework specially focused

to OLAP analysis (rather than the underlying

aggregate querying layer), which indeed defines a

powerful reference application scenario for a wide

spectrum of emerging applications over data streams,

such as security (e.g., (Zhou et al., 2014; R.H. Deng

et al., 2014)) and emergency management (e.g., (Gan

& Dai, 2014; Keren et al., 2014)).

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 multi-step 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

282

Cuzzocrea, A., Schuster, A. and Vercelli, G.

PP-OMDS: An Effective and Efﬁcient Framework for Supporting Privacy-Preserving OLAP-based Monitoring of Data Streams.

DOI: 10.5220/0006812102820292

In Proceedings of the 20th International Conference on Enterprise Information Systems (ICEIS 2018), pages 282-292

ISBN: 978-989-758-298-1

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 PP-

OMDS 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 trade-off

analysis.

2 RELATED WORK

In this Section, we first review the scientific and

technological background of the three distinct PP-

OMDS components, namely: privacy-preserving

OLAP, data stream monitoring, and privacy-

preserving data stream monitoring.

Privacy-preserving OLAP over data streams is a

research challenge that puts down roots in the basic

privacy-preserving OLAP problem over static (e.g.,

relational) data (e.g., (Cuzzocrea et al., 2008;

Cuzzocrea & Saccà, 2012)). Basically, the latter

problem origins from recognizing that malicious

users can infer sensitive knowledge from online

corporate data cubes that do not adopt effective

privacy preserving countermeasures (e.g., (Wang et

al., 2004a; Wang et al., 2004b). From this breaking

evidence, which originally derives from the privacy-

preserving database management problem (e.g.,

(Sweeney, 2002; Machanavajjhala et al., 2007)), a

plethora of Privacy Preserving Data Mining (PPDM)

(Agrawal et al., 2000) techniques has been proposed

during the last years. Each of these techniques focuses

on supporting the privacy preservation of a

specialized KDD/DM task such as frequent item set

mining (e.g., (Li et al., 2012)), clustering (e.g.,

(Alotaibi et al., 2014)), and so forth. Privacy-

preserving OLAP (Agrawal, et al., 2005) is a specific

PPDM technique dealing with the privacy

preservation of data cubes (Gray et al., 1997). Data

cubes play a leading role in Data Warehousing (DW)

and Business Intelligence (BI) systems, as, on the

basis of a multidimensional and a multi-resolution

vision of data, data cubes make available to OLAP

users/applications SQL aggregations (e.g., SUM,

COUNT, AVG etc.) computed over very large

amounts of data stored in external sources (e.g.,

relational databases). These aggregations enable

OLAP users/applications to easily extract

summarized knowledge from the underlying massive

data sources, with performance infeasible for

traditional OLTP processes (e.g., (Cuzzocrea et al.,

2009; Cuzzocrea & Matrangolo, 2004)).

Unfortunately, as highlighted by recent studies

(Pernul et al., 2000; Wang et al., 2004a; Wang et al.,

2004b; Agrawal et al., 2005; Hua et al., 2005; S.Y.

Sung et al., 2006), the privacy risk heavily affects

online-published data cubes. By accessing and

querying data cubes, malicious users can infer OLAP

aggregations computed over sensitive ranges of

multidimensional data that, due to privacy reasons,

are hidden to unauthorized users. Specifically, since

OLAP deals with aggregate data and summarized

knowledge, malicious users are usually interested in

PP-OMDS: An Effective and Efﬁcient Framework for Supporting Privacy-Preserving OLAP-based Monitoring of Data Streams

283

inferring what we define as aggregate patterns of

multidimensional data, rather than individual

information of data cells stored in data cubes (e.g.,

(Sung et al., 2006)) or tuples stored in relational

databases (e.g., (Sweeney, 2002; Machanavajjhala et

al., 2007)). Hence, the final goal of privacy-

preserving OLAP techniques consists in protecting

the privacy of aggregate patterns of sensitive interest

for the target organization.

Directly deriving from the static case, the dynamic

case, i.e. OLAP over data streams (e.g., (Cuzzocrea,

2013; Cuzzocrea, 2011; Cuzzocrea, 2009)), extends

the basic problem to the more challenging issue of

dealing with data streams (e.g., (Babcock et al.,

2006)). This introduces harder research challenges to

be considered, due to the fact that additional

requirements of making privacy-preserving OLAP

aggregations computed over data streams occur (e.g.,

(Cuzzocrea, 2009)), mainly coming from the well-

known constraints dictated by data stream processing,

i.e. bounded memory, single-pass algorithms, near-

real-time computation (e.g., (Golab & M.T. Özsu,

2003)). Looking at the actual literature, it follows

that, while there is considerable amount of research

work on the problem of effectively and efficiently

executing Data Mining tasks over data streams such

as clustering and classification (e.g., (Gaber et al.,

2005; C.C. Aggarwal & P.S. Yu, 2008; Zhang et al.,

2008)), there exist in literature few research

initiatives that focus the attention on the yet-relevant

issue of effectively and efficiently supporting OLAP

over data streams (e.g., (Cuzzocrea, 2009)), also

taking into consideration uncertainty and imprecision

that very often characterized such data type (e.g.,

(Cuzzocrea, 2013; Cuzzocrea, 2011)), which,

similarly (as to confirm their relevance), have also

been studied in well-known Data Mining settings

(e.g., (Cuzzocrea et al., 2014)). (Cormode &

Garofalakis, 2007) and (Jayram et al., 2007) are

studies that put theoretical foundations to the OLAP

over data streams problem, since they focus on the

problem of computing aggregates over probabilistic

data streams. Indeed, probabilistic data stream

models, which have been firstly introduced in (L.J.C.

Re et al., 2010), are well suited to represent inherent

characteristics of data streams. Unfortunately, these

approaches do not focus on specific features of

streaming data, such as multidimensionality

(Cuzzocrea, 2009), which is indeed critical in data

stream analysis (Cai, 2004; Han et al., 2005). In line

with this research initiative, (Cormode & Garofalakis,

2007) and (Jayram et al., 2007) offer theoretical

models for understanding and capturing aggregations

in probabilistic data streams, mainly via the idea of

exploiting Probability Distribution Function (PDF)

to model uncertain and imprecise data streams. (M.-

J. Hsieh et al., 2007) has some relation with the

research related to OLAP over data streams, due to

the fact it studies the issue of compressing stream

cubes by inheriting well-known models and

algorithms for approximate query processing, with

the idea of adding more efficiency and reliability.

Lahar (L.J.C. Re et al., 2010) is a specialized

warehousing system that focuses on so-called

Markovian streams, which well model and capture

uncertainty and imprecision in streaming data, and

proposes approximate methods for answering event-

OLAP queries over such class of data streams via

trading-off efficiency and accuracy. (Jampani et al.,

2008) is also relevant to our research due to the fact it

introduces a nice Monte-Carlo-based approach for

managing uncertain data that could be easily

incorporated in our proposed PP-OMDS framework

in order to find finer probabilistic models.

Data stream monitoring is an active area of

research where the main issue consists in devising

models, techniques and algorithms for supporting the

online monitoring of streaming data, which is

relevant for a wide spectrum of applications ranging

from telecommunication networks to web-click

streams, from intrusion detection systems to sensor

networks, and so forth. As highlighted by recent

studies (e.g., (Bulut, et al., 2009)), the main research

challenge to deal-with in this context is represented

by the strong need for efficiency in terms of space

usage and per-item processing time while providing

high-accuracy answers to several classes of useful

queries, such as aggregate monitoring queries, aimed

at finding surprising levels in the target data stream,

and similarity queries, devoted to detect correlations

and find interesting patterns in the target data stream.

This research line has originated a rich collection of

proposals that focus the attention on the specific data

stream monitoring problem. (Gan & Dai, 2014)

proposes a technique for detecting dynamic changes

in frequencies of episodes over time-evolving

streams, with special emphasis on online detection of

emerging patterns such as abrupt emerging episodes

and abrupt submerging episodes (over streams). (K.-

Y. Cao, et al., 2014) proposes Continuous Uncertain

Outlier Detection (CUOD), a framework capable of

quickly determining the nature of uncertain elements

in data streams, by exploiting pruning strategies in

order to improve the efficiency. Furthermore, a

further alternative pruning approach, named as

Probability Pruning for Continuous Uncertain Outlier

Detection (PCUOD), is proposed to reduce the

detection cost. The latter is a method for

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

284

probabilistically estimating outliers in streams, in

order to effectively reduce the amount of

computational overhead. Communication-efficient

monitoring of distributed streams has also been the

subject of much research in recent years. Some

research has focused on anomaly detection (Huang et

al., 2006; Huang et al., 2007), while other studies

focus on monitoring specific types of functions,

including sums (Keralapura et al., 2006; Olston et al.,

2003), Boolean predicates (Agrawal et al., 2006),

inner products (Cormode & M.N. Garofalakis, 2008)

and entropy (Arackaparambil et al., 2008). (Sharfman

et al., 2006) employs techniques presented in the

context of geometric monitoring, which enable

monitoring arbitrary threshold functions by

interpreting the monitoring task as a geometric

problem. This research has been important extensions

recently (e.g., (Lazerson et al., 2015; Keren et al.,

2014; Giatrakos et al., 2014; Keren et al., 2012)).

Privacy-preserving data stream monitoring is

directly related to our research (Keren et al., 2014;

Friedman et al., 2014). The application of differential

privacy (Dwork, 2006) to data stream processing has

been studied initially in (Dwork et al., 2010), which

has introduced the concept of pan-private data stream

algorithms – algorithms that retain their privacy

properties even when intrusions ex-pose the internal

state of the system. Two independent works (Dwork

et al., 2010b) and (T.-H.H. Chan et al., 2010) study

continuous release of differentially-private counts,

optionally while ensuring pan-privacy. Mir et al. (Mir

et al., 2011) rely on sketches to track statistics such

as distinct count, cropped first moment, and heavy

hitters count over fully dynamic data while preserving

pan-privacy. While we do not aim to obtain pan-

privacy, the PP-OMDS framework could be extended

to support it through straightforward application of

the technique of (Dwork et al., 2010b). Indeed,

exploiting differential-privacy tools (e.g., (Dwork,

2006a)), combined with OLAP over data streams

(e.g., (Cuzzocrea, 2009)), with monitoring goals (e.g.,

(Keren et al., 2014; Friedman et al., 2014)) is very

promising for achieving the PP-OMDS project’s

goals. Early works on differential privacy in a

distributed setting (Dwork et al., 2006, Beimel et al.,

2008) study how differential privacy could be

combined with cryptographic protocols to allow one-

off computations to be carried out both securely and

privately. Chen et al. (Chen et al., 2012) make use of

the noise generation mechanism of (Dwork et al.,

2006) to allow analysts to pose histogram queries to

a subset of distributed clients with the help of an

honest but curious proxy. (Friedman et al., 2014)

proposes a general framework that enables

monitoring arbitrary functions over statistics derived

from distributed data stream in a privacy-preserving

manner.

Several works study differentially-private

aggregation over distributed time-series data,

focusing mostly on simple aggregates, such as counts

and sums. Rastogi and Nath (Rastogi et al., 2010) rely

on the Discrete Fourier Transform to compress

historical time-series data, and leverage threshold

homomorphic encryption to run a distributed version

of the Laplace mechanism on the compressed data. As

the compression requires access to all the query

results in advance, this method is not adequate for

processing data streams on the fly. Shi et al. (Shi et

al., 2011) apply cryptographic techniques to allow an

untrusted aggregator to compute differentially-

private sums over distributed peers without learning

anything but the outcome. While the proposed

scheme has been designed to reduce the overhead of

cryptographic operations in periodical

communications, it does not address the cumulative

privacy loss. Fan and Xiong (Fan & Xiong, 2012)

address the dynamic nature of the data by adaptive

sampling of the time-series data and use of Kalman

filters for estimating the data in non-sampling points.

3 REFERENCE APPLICATION

SCENARIO

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

(Zhou et al., 2014; R.H. Deng et al., 2014),

emergency management (Gan & Dai, 2014; Keren et

al., 2014), and so forth). Figure 1 shows a typical

application scenario for the PP-OMDS framework.

Here, a sensor network, composed by both sink nodes

and sensor nodes (e.g., (Cuzzocrea, 2014)), produces

sensor readings of kind



  



such that: (i)  is

the absolute identifier of the sensor node, (ii)  is the

timestamp at which the sensor reading is produced,

(iii)  is the proper reading (i.e., the value). A

privacy-preserving 2D OLAP-based monitoring view

 is interfaced to the sensor network directly, and it

is used to monitor the reading value variable , based

PP-OMDS: An Effective and Efﬁcient Framework for Supporting Privacy-Preserving OLAP-based Monitoring of Data Streams

285

Figure 1: A Typical Application Scenario for the PP-OMDS Project.

on a complex multidimensional data model (Gray et

al., 1997; Cuzzocrea, 2009). In the reference

application scenario of Figure 1, the model introduces

 as the first dimension and  as the

second dimension, respectively, and the measure

value is defined on top of a COUNT aggregate

operator over a two-dimensional range







 





, being





and 



two one-dimensional ranges along

 and , respectively, which define the

data cube cell. The variable is monitored via suitable

aggregate monitoring queries (e.g., (Bulut, et al.,

2009)) over data cube cells that populate the OLAP

view .

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 privacy-

preserving manner (e.g., (Cuzzocrea et al., 2008;

Cuzzocrea & Saccà, 2012) – for the static case), i.e. it

must preserve the privacy of data sources that

originate the sensor readings. The OLAP view  is

meant for monitoring goals: when the COUNT-

aggregated value of a data cube cell 



exceeds a fixed

threshold , then an event occurs and a suitable action

(trigger, respectively) is activated in order to modify

the target monitored environment. Obviously, the

described application scenario is just an instance that

represents even more sophisticated multidimensional

settings, where the privacy-preserving OLAP-based

monitoring view is characterized by multiple

dimensions (e.g., (Han et al., 2005)). It is worth to

recognize that the described application scenario

well-describes a plethora of modern data stream

applications, ranging from event detection (e.g., (Gyu

Kim & Kim, 2014)) to complex monitoring queries

support (e.g., (Li, 2014)), from near-duplicate

detection over multimedia streams (e.g., (C.-Yi Chiu

et al., 2013)) to anomaly detection (e.g., (Taylor et

al., 2013)), and so forth, all with the innovative and

challenging requirement of preserving the privacy of

data streams (e.g., (Al-Hussaeni et al., 2014; Kim et

al., 2014)), under OLAP analysis requirements (e.g.,

(Cuzzocrea, 2013; Cuzzocrea, 2011)), with

monitoring purposes (e.g., (Keren et al., 2014;

Friedman et al., 2014)), being the latter the three main

pillar of the PP-OMDS project.

4 IMPACT AND EXPLOITATIONS

The PP-OMDS framework provides 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” carried-out

by the framework.

As highlighted in Section 1 and Section 2, where

research topics of the PP-OMDS framework, along

with the background state-of-the-art work, have been

described in details, this will represent a critical

advancement of actual literature, because while

several proposals on privacy-preserving data stream

monitoring exist (e.g., (Dwork et al., 2010b; T.-H.H.

Chan et al., 2010; D.J. Mir et al., 2011; Dwork et al.,

2006; Beimel et al., 2008; Chen et al., 2012;

Friedman et al., 2014)), as well as several proposals

on privacy-preserving OLAP, the majority on static

data (e.g., (Cuzzocrea et al., 2008, Cuzzocrea &

Saccà, 2012)), and few proposal of OLAP over

streaming data (e.g., (Cuzzocrea, 2013; Cuzzocrea,

2011; Cuzzocrea, 2009)), research efforts have not

investigated, till now, the yet relevant challenge of

combining the basic components under the vision of

a common framework specially focused to privacy-

preserving OLAP-based monitoring of data streams.

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

286

Along with the models, techniques and algorithms

above, the PP-OMDS framework also outcomes other

contributions, such as:

 theoretical complexity analysis (asymptotic

issues and possible lower bounds);

 properties and lemma deriving from the main

theory;

 computational paradigms;

 reference case studies and application

scenarios;

 performance analysis of devised algorithms;

 robustness and sensitivity analysis of devised

algorithms;

 optimization strategies and solutions for

devised algorithms.

Given the specificities of the investigated

problem, even theoretical privacy-analysis-oriented

contributions are expected. These comprise:

 theoretical analysis of the privacy-preserving

capabilities of devised models and

techniques;

 analysis and detection of possible privacy

breaches (e.g., (Chi-Wing Wong et al.,

2011));

 analysis of the robustness of devised models

and techniques against harder attack

configurations such as coalitions (e.g.,

(Cuzzocrea & Bertino, 2011)).

All these contributions clearly represent a

milestone for future research efforts in the field,

which can be reasonable intended as one of the

fundamental expected impacts of the PP-OMDS

framework. Moreover, as already highlighted, the PP-

OMDS framework is “naturally” compliant with the

topics carried-on within the EU H2020 research

framework, as they essentially investigate the topic

Big Data, which is very relevant within H2020 with a

relevant number of calls focused on it in all the main

(H2020) axes, i.e. Excellence Science, Industrial

Leadership, Societal Challenge.

In addition to the main research-oriented impact,

the PP-OMDS framework also provides a number of

technological expected outcomes, which can be

summarized as reference technical and algorithmic

solutions for effectively and efficiently supporting

privacy-preserving OLAP-based monitoring of data

streams in critical data-intensive settings, such as:

 distributed processing of streaming financial

data in inter-organizational settings (e.g., like

the ones built upon emerging Cloud

Computing infrastructures) – in this

application scenario, participants (e.g., bank,

insurance, government agency, and so forth)

would like to disclose aggregate information

for inter-organizational processing support

(e.g., monitoring pre-defined balance

thresholds) while, at the same, still protecting

aggregates over their own sensitive ranges of

data;

 distributed processing of streaming social

inter-network data for analysis purposes – in

this application scenario, streaming data

coming from social inter-networks are

monitored and analyzed according to OLAP-

based multidimensional models in order to

discover useful knowledge (e.g., anomalies,

alarms, emerging patterns, user profiling and

user tracking metrics, and so forth) focused

to the interaction of different social networks,

even defined on top of heterogeneous

semantic domains.

5 PP-OMDS ARCHITECTURE

AND MAIN

FUNCTIONALITIES

Figure 2 shows the detailed architecture of the PP-

OMDS framework, along with its components and

reference technologies. These are:

 Data Stream Source Layer – The layer where

data streams are produced by a target process

(e.g., sensor networks, logistic networks,

etc.). Technology used: arbitrary (at the input

layer).

 Data Stream Acquisition Server – The server

devoted to acquire data streams (by also

providing the initial pre-processing – e.g.,

cleaning) and to make the necessary data

model transformations. Technology used:

Java programming language; JBoss

application server, MySQL database server.

In particular: (i) JBoss provides the necessary

Java execution runtime environment for the

Data Stream Collection Component, which is

the component that implements classes and

functions supporting the data 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

PP-OMDS: An Effective and Efﬁcient Framework for Supporting Privacy-Preserving OLAP-based Monitoring of Data Streams

287

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

Figure 2: PP-OMDS Framework Detailed Architecture.

Management Component, which is the

component that implements classes and

functions supporting the privacy-preserving

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 PP-OMDS 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).

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

288

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 PP-OMDS 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).

Table 1: Comparative Analysis of Competitor Frameworks and the PP-OMDS Framework.

Distributed Data

Stream Monitoring

OLAP Analysis over

Data Streams

Privacy-Preserving

Data Stream

Management

Support for General-

Purpose Applications

Support for Vertical

Applications

APAMA Streaming

Analytics









Stream Computing









Event Stream

Processor









Apache Spark









PP-OMDS



6 COMPARATIVE ANALYSIS

WITH COMPETITIVE

FRAMEWORKS

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)

PP-OMDS: An Effective and Efﬁcient Framework for Supporting Privacy-Preserving OLAP-based Monitoring of Data Streams

289

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 PP-

OMDS framework’s paradigms. Table 1 proposes a

summary on the comparative analysis of the

competitor frameworks against the PP-OMDS

framework, along the above-discussed

parameters/functionalities.

7 CONCLUSIONS

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.

REFERENCES

A. Beimel, K. Nissim, E. Omri. Distributed private data

analysis: Simultaneously solving how and what,

CRYPTO, pp. 451-468, 2008

A. Bulut, N. Koudas, A. Meka, A.K. Singh, D. Srivastava:

Optimization Techniques for Reactive Network

Monitoring. IEEE Trans. Knowl. Data Eng. 21(9):

1343-1357, 2009

A. Cuzzocrea, C.K.-S. Leung, R.K. MacKinnon: Mining

constrained frequent itemsets from distributed

uncertain data. Future Generation Computer Systems

37: 117-126, 2014

A. Cuzzocrea, D. Saccà: A Theoretically-Sound

Accuracy/Privacy-Constrained Framework for

Computing Privacy Preserving Data Cubes in OLAP

Environments. OTM Conferences, pp. 527-548, 2012

A. Cuzzocrea, E. Bertino: Privacy Preserving OLAP over

Distributed XML Data: A Theoretically-Sound Secure-

Multiparty-Computation Approach. Journal of

Computer and System Sciences 77(6): 965-987, 2011

A. Cuzzocrea, F. Furfaro, D. Saccà: Enabling OLAP in

mobile environments via intelligent data cube

compression techniques. J. Intell. Inf. Syst. 33(2): 95-

143 2009

A. Cuzzocrea, U. Matrangolo: Analytical Synopses for

Approximate Query Answering in OLAP

Environments, DEXA, 2004

A. Cuzzocrea, V. Russo, D. Saccà: A Robust Sampling-

Based Framework for Privacy Preserving OLAP.

DaWaK, 2008, pp. 97-114

A. Cuzzocrea: Approximate OLAP Query Processing over

Uncertain and Imprecise Multidimensional Data

Streams. DEXA, pp. 156-173, 2013

A. Cuzzocrea: CAMS: OLAPing Multidimensional Data

Streams Efficiently. DaWaK, pp. 48-62, 2009

A. Cuzzocrea: Optimization issues of querying and

evolving sensor and stream databases. Inf. Syst. 39:196-

198, 2014

A. Cuzzocrea: Retrieving Accurate Estimates to OLAP

Queries over Uncertain and Imprecise

Multidimensional Data Streams. SSDBM, 2011, pp.

575-576

A. Friedman, I. Sharfman, D. Keren, A. Schuster: Privacy-

Preserving Distributed Stream Monitoring, NDSS, 2014

A. Lazerson, I. Sharfman, D. Keren, A. Schuster, M.N.

Garofalakis, V. Samoladas: Monitoring Distributed

Streams using Convex Decompositions. PVLDB 8(5):

545-556, 2015

A. Machanavajjhala, et al. L-diversity: Privacy beyond k-

Anonymity. ACM Trans. on Knowledge Discovery

from Data, 1(1): art. no. 3, 2007

B. Babcock, S. Babu, M. Datar, R. Motwani, J. Widom:

Models and Issues in Data Stream Systems, PODS,

2002, pp. 1-16

C. Arackaparambil, J. Brody, A. Chakrabarti: Functional

monitoring without monotonicity. ICALP, 2009

C. Dwork, K. Kenthapadi, F. Mcsherry, I. Mironov, M.

Naor: Our data, ourselves: Privacy via distributed noise

generation, EUROCRYPT, 2006

C. Dwork, M. Naor, T. Pitassi, G.N. Rothblum, S.

Yekhanin: Pan-private streaming algorithms, ICS, 2010

C. Dwork, T. Pitassi, M. Naor, G.N. Rothblum: Differential

privacy under continual observation, STOC, 2010

C. Dwork: Differential Privacy, ICALP (2) 2006, pp. 1-12

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

290

C. Olston, J. Jiang, J. Widom: Adaptive filters for

continuous queries over distributed data streams,

SIGMOD, 2003

C.-Yi Chiu, T.-H. Tsai, G.-W. Han, C.-Y. Hsieh, S.-Y. Li:

Efficient Video Stream Monitoring for Near-Duplicate

Detection and Localization in a Large-Scale

Repository. ACM Trans. Inf. Syst. 31(4):22, 2013

C.C. Aggarwal, P.S. Yu: A Framework for Clustering

Uncertain Data Streams, IEEE ICDE, 2008

D. Keren, G. Sagy, A. Abboud, D. Ben-David, A. Schuster,

I. Sharfman, A. Deligiannakis: Geometric Monitoring

of Heterogeneous Streams. IEEE Trans. Knowl. Data

Eng. 26(8): 1890-1903, 2014

D. Keren, G. Sagy, A. Abboud, D. Ben-David, A. Schuster,

I. Sharfman, A. Deligiannakis: Monitoring Distributed,

Heterogeneous Data Streams: The Emergence of Safe

Zones. ICAA, 2014, pp. 17-28

D. Keren, I. Sharfman, A. Schuster, A. Livne: Shape

Sensitive Geometric Monitoring. IEEE Trans. Knowl.

Data Eng. 24(8): 1520-1535, 2012

D.J. Mir, S. Muthukrishnan, A. Nikolov, R.N. Wright: Pan-

private algorithms via statistics on sketches, PODS,

2011, pp. 37-48

E. Shi, T.-H. H. Chan, E.G. Rieffel, R. Chow, D. Song:

Privacy-preserving aggregation of time-series data,

NDSS, 2011

G. Cormode, M. Garofalakis: Sketching Probabilistic Data

Streams, ACM SIGMOD, 2007

G. Cormode, M.N. Garofalakis: Approximate continuous

querying over distributed streams. ACM Transactions

on Database Systems, 33(2), 2008

G. Li, C. Luo, J. Li: Continuous Monitoring of Top-k

Dominating Queries over Uncertain Data Streams.

WISE, 2014, pp. 244-255

G. Pernul, et al. Towards OLAP Security Design - Survey

and Research Issues, ACM DOLAP, pp. 114-121, 2000

Gray J., et al. Data Cube: A Relational Aggregation

Operator Generalizing Group-By, Cross-Tab, and Sub-

Totals. Data Mining and Knowledge Discovery, 1(1):

29-54, 1997

H. Gyu Kim, C. Kim: Interval clustering algorithm for fast

event detection in stream monitoring applications.

Pattern Recognition Letters 36:171-176, 2014

I. Sharfman, A. Schuster, D. Keren: A geometric approach

to monitoring threshold functions over distributed data

streams, SIGMOD, 2006

I. Taylor, J.L. Sharp, D.L. White, J.O. Hallstrom, G.W.

Eidson, J.B. von Oehsen, E.B. Duffy, C.V. Privette III,

C.T. Cook, A. Sampath, G. Radhakrishnan: Monitoring

Sensor Measurement Anomalies of Streaming

Environmental Data Using a Local Correlation Score.

COM.Geo, 2013, pp.136-137

J. Han, Y. Chen, G. Dong, J. Pei, B.W. Wah, J. Wang, Y.D.

Cai: Stream Cube: An Architecture for Multi-

Dimensional Analysis of Data Streams, Distributed and

Parallel Databases 18(2), 2005

J. Smailovic, M. Grcar, N. Lavrac, M. Znidarsic: Stream-

based active learning for sentiment analysis in the

financial domain. Inf. Sci. 285:181-203, 2014

J. Zhou, O.C. Au, G. Zhai, Y.Y. Tang, X. Liu: Scalable

Compression of Stream Cipher Encrypted Images

Through Context-Adaptive Sampling. IEEE

Transactions on Information Forensics and Security

9(11):1857-1868, 2014

K. Al-Hussaeni, B.C.M. Fung, W.K. Cheung: Privacy-

preserving trajectory stream publishing. Data Knowl.

Eng. 94:89-109, 2014

K. Alotaibi, V.J. Rayward-Smith, B. de la Iglesia:

Nonmetric multidimensional scaling: A perturbation

model for privacy-preserving data clustering. Statistical

Analysis and Data Mining 7(3):175-193, 2014

K.-Y. Cao, G.-R. Wang, D.-H. Han, G.-H. Ding, A.-X.

Wang, L.-X. Shi: Continuous Outlier Monitoring on

Uncertain Data Streams. J. Comput. Sci. Technol.

29(3):436-448, 2014

L. Fan, L. Xiong. Real-time aggregate monitoring with

differential privacy, CIKM, 2012

L. Golab, M.T. Özsu: Issues in data stream management.

SIGMOD Record 32(2): 5-14, 2003

L. Huang, M. Garofalakis, J. Hellerstein, A. Joseph, and N.

Taft. Toward sophisticated detection with distributed

triggers, SIGCOMM, 2006

L. Huang, X. L. Nguyen, M. Garofalakis, J. M. Hellerstein,

M. I. Jordan, A. D. Joseph, and N. Taft.

Communication-efficient online detection of network-

wide anomalies, INFOCOM, 2007

L. Sweeney, k-Anonymity: A Model for Protecting Privacy.

Int. Jou. on Uncertainty Fuzziness and Knowledge-

based Systems, 10(5): 557-570, 2002

L. Wang, et al. Cardinality-based Inference Control in Data

Cubes. Jou. of Computer Security, 12(5): 655-692,

2004

L. Wang, et al. Securing OLAP Data Cubes against Privacy

Breaches, IEEE SSP, pp. 161-175, 2004

L.J.C. Re, M. Balazinska, M. Philipose: Approximation

Trade-Offs in Markovian Stream Processing: An

Empirical Study, IEEE ICDE, 2010

M. Gaber, A. Zaslavsky, S. Krishnaswamy: Mining Data

Streams: A Review, SIGMOD Record 34(2), 2005

M. Gan, H. Dai: Detecting and monitoring abrupt

emergences and submergences of episodes over data

streams. Inf. Syst. 39:277-289, 2014

M. Hua, et al. FMC: An Approach for Privacy Preserving

OLAP, DaWaK, pp. 408-417, 2005

M.-J. Hsieh, M.-S. Chen, P.S. Yu: Approximate Query

Processing in Cube Streams, IEEE Trans. on

Knowledge and Data Engineering 19(11), 2007

N. Giatrakos, A. Deligiannakis, M.N. Garofalakis, I.

Sharfman, A. Schuster: Distributed Geometric Query

Monitoring Using Prediction Models. ACM Trans.

Database Syst. 39(2): 16, 2014

N. Li, W.H. Qardaji, D. Su, J. Cao: PrivBasis: Frequent

Itemset Mining with Differential Privacy. PVLDB

5(11):1340-1351, 2012

Q. Zhang, F. Li, K. Yi: Finding Frequent Items in

Probabilistic Data, ACM SIGMOD, 2008

R. Agrawal, et al. Privacy-Preserving Data Mining, ACM

SIGMOD, pp. 439-450, 2000

PP-OMDS: An Effective and Efﬁcient Framework for Supporting Privacy-Preserving OLAP-based Monitoring of Data Streams

291

R. Agrawal, et al. Privacy-Preserving OLAP, ACM

SIGMOD, pp. 251-262, 2005

R. Chen, A. Reznichenko, P. Francis, J. Gehrke: Towards

statistical queries over distributed private user data,

NSDI, 2012

R. Chi-Wing Wong, A. Wai-Chee Fu, K. Wang, P.S. Yu,

Jian Pei: Can the Utility of Anonymized Data be Used

for Privacy Breaches? ACM Trans. on Knowledge

Discovery from Data 5(3):16, 2011

R. Jampani, F. Xu, M. Wu, L.L. Perez, C. Jermaine, P.J.

Haas: MCDB: A Monte Carlo Approach to Managing

Uncertain Data, ACM SIGMOD, 2008

R. Keralapura, G. Cormode, J. Ramamirtham:

Communication-efficient distributed monitoring of

thresholded counts, SIGMOD, 2006

R.H. Deng, X. Ding, S.-W. Lo: Efficient authentication and

access control of scalable multimedia streams over

packet-lossy networks. Security and Communication

Networks 7(3):611-625, 2014

S. Agrawal, S. Deb, K. Naidu, R. Rastogi: Efficient

detection of distributed constraint violations, ICDE,

2006

S. Kim, M.K. Sung, Y.D. Chung: A framework to preserve

the privacy of electronic health data streams. Journal of

Biomedical Informatics 50:95-106, 2014

S.Y. Sung, et al. Privacy Preservation for Data Cubes.

Knowledge and Information Systems, 9(1): 38-61, 2006

T.-H.H. Chan, E. Shi, D. Song: Private and continual

release of statistics, ICALP, 2010

T.S. Jayram, A. McGregor, S. Muthukrishnan, E. Vee:

Estimating Statistical Aggregates on Probabilistic Data

Streams, ACM PODS, 2007

V. Rastogi, S. Nath: Differentially private aggregation of

distributed time-series with transformation and

encryption, SIGMOD, 2010

Y.D. Cai, D. Clutterx, G. Papex, J. Han, M. Welgex, L.

Auvilx: MAIDS: Mining Alarming Incidents from Data

Streams, ACM SIGMOD, 2004

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

292