Skip to main content

Alina Campan

Followers

20

Following

6

Co-authors

6

Public Views

Professor of Computer Science at NKU.

less

London South Bank University

University of Cologne

Roshan Chitrakar

Nepal College of Information Technology

Graduate Center of the City University of New York

Alejandro Trelles

Brandeis University

The University Of Jordan

Prof. Hena Vadi

GTU

Christopher Crick

Oklahoma State University

Interests

Uploads

Papers by Alina Campan

Search and Retrieval in Semantic-Structural Representations of Novel Malware

Advances in Artificial Intelligence and Machine Learning, 2024

In this study, we present a novel representation for binary programs which captures semantic simi... more In this study, we present a novel representation for binary programs which captures semantic similarity and structural properties. This representation enables the search and retrieval of binary executable programs based on their similarity of behavioral properties. The proposed representation is composed in a bottom-up approach: we begin by extracting data dependency graphs (DDG), which are representative of both program structure and operational semantics. We then encode each program as a set of graph hashes representing isomorphic uniqueness, a method we have labeled DDG Fingerprinting. We present experimental results of search using k-Nearest Neighbors in a metric space constructed from a set of binary executables. Searches in the dataset are based on the operational semantics of specific malware examples. 2052

Sensitive Attributes based Privacy Preserving in Data Mining using k-anonymity

knowledge from huge amount of data. In recent years, there has been a tremendous growth in the am... more knowledge from huge amount of data. In recent years, there has been a tremendous growth in the amount of personal data that can be collected and analyzed by the organizations. Organizations such as credit card companies, real estate companies and hospitals collect and hold large volumes of data for their research purposes. E.g. National Institute of health. When these organizations publish data containing a lot of sensitive information. The importance of sharing data for research and knowledge discovery has been well-recognized. However, sharing data that contains sensitive personal information, such as insurance data, medical record, etc across organization boundaries can raise serious privacy concerns. There is a need to preserve the privacy of the individuals in data set . K-anonymity is one of the easy and efficient techniques to achieve privacy in many data publishing applications. In k-anonymity, all tuples of releasing database are generalized to make it anonymize which lead to data utility reduction and more information loss of publishing table. Sensitive attribute based anonymity method is very useful in preserving the privacy of individuals in organization’s publication of data. It reduces information loss to the researchers by providing sensitive levels. This method also avoids Homogeneity attack and Background attacks.

Efficient Approximation Algorithms for Minimum Dominating Sets in Social Networks

International Journal of Service Science, Management, Engineering, and Technology, Apr 1, 2018

Socialnetworksareincreasinglybecominganoutletthatismoreandmorepowerfulinspreading newsandinfluenc... more Socialnetworksareincreasinglybecominganoutletthatismoreandmorepowerfulinspreading newsandinfluenceindividuals.Comparedwithothertraditionalmediaoutletssuchasnewspaper, radio,andtelevision,socialnetworksempoweruserstospreadtheirideologicalmessageand/orto delivertargetadvertisingveryefficientlyintermsofbothcostandtime.Inthisarticle,theauthors focusonefficientlyfindingdominatingsetsinsocialnetworksfortheclassicaldominatingsetproblem aswellasfortworelatedproblems:partialdominatingsetsandd-hopdominatingsets.Theywill presentalgorithmsfordeterminingefficientlyagoodapproximationforthesocialnetworkminimum dominatingsetsforeachofthethreevariants.Theauthorsrananextensivesuiteofexperimentsto testthepresentedalgorithmsonseveraldatasetsthatincluderealnetworksmadeavailablebythe StanfordNetworkAnalysisProjectandsyntheticnetworksthatfollowthepower-lawandrandom modelsthattheygeneratedforthiswork.Theperformedexperimentsshowthattheselectionofthe algorithmthatperformsbesttodetermineefficientlythedominatingsetisdependentofnetwork characteristics and the order of importance between the size of the dominating set and the time requiredtodeterminesuchaset.

Exploring Sentiment on Campus. A Twitter Sentiment Analysis on University Tweets

EPiC series in computing, Mar 21, 2024

We report in this work the results of our analysis of accuracy of 5 sentiment analysis methods (T... more We report in this work the results of our analysis of accuracy of 5 sentiment analysis methods (TextBlob, VADER, logistic regression, support vector machine, CNN on encodings based on BERT tokenization,) for a dataset consisting of tweets from the academia domain, that we API-scraped for 32 universities during the year 2022. We show some results for the volume and sentiment polarity trends exhibited by this dataset. We connect peak and low sentiment averages to concrete events that explain the respective sentiment trend; this proves that observing the social media trends allows to detect real events that need attention and possible action.

A new approach for optimal fragmentation order in distributed object oriented databases

Pure mathematics and applications, 2004

Class fragmentation is an important task in the design of Distributed OODBs and there are many al... more Class fragmentation is an important task in the design of Distributed OODBs and there are many algorithms handling it. Almost none of them deals however with the class fragmentation order details. We claim that class fragmentation order can induce severe performance penalties if not considered in the fragmentation phase. We propose here two variants of an algorithm for finding the optimal class fragmentation order in a DOODB. In both variants we capture all class relations (inheritance, aggregation) and we determine a class fragmentation order where precedence conflicts induced by relation cycles are eliminated in such way that the strongest links be maintained.

A comparison between local and global recoding algorithms for achieving microdata -sensitive -anonymity

Acta Universitatis Apulensis. Mathematics - Informatics, 2008

New privacy regulations together with ever-increasing data availability and computational power h... more New privacy regulations together with ever-increasing data availability and computational power have created a huge interest in data privacy research. One major research direction is built around k-anonymity property, which is required for the released data. Although k-anonymity protects against identity disclosure, it fails to provide an adequate level of protection with respect to attribute disclosure. We introduced a new privacy protection property called p-sensitive k-anonymity that avoids this shortcoming. We developed new algorithms (GreedyPKClustering and EnhancedPKClustering) and adapted an existing algorithm (Incognito) to generate masked microdata with p-sensitive k-anonymity property. All these algorithms try to reduce the amount of information lost while transforming data to conform to p-sensitive k-anonymity. They are different in the masking methods they use. The new algorithms are based on local recoding masking methods. Incognito, initially designed for k-anonymity, uses global recoding for masking. This paper's goal is to compare the impact of the masking method on the quality of the masked microdata obtained. For this we compare the quality of the results (cost measures based on data utility) and the efficiency (running time) of these three algorithms for masking both real and synthetic data sets.

Efficient approximation algorithms to determine minimum partial dominating sets in social networks

In this paper we report on extensive experiments for determining partial dominating sets of small... more In this paper we report on extensive experiments for determining partial dominating sets of small size for various types of real and synthetic social networks. Our experiments ran on several real network datasets made available by the Stanford Network Analysis Project and on some synthetic power-law and random networks created with social network generators. To compute partial dominating sets on these networks we used five algorithms compared in [4], which were adapted for partial dominating sets. Our experiments showed that there are several good algorithms that can efficiently find quality approximations for the minimum-size partial dominating set problem. The best algorithm choice is dependent on the network characteristics and the value of the coverage parameter.

P-Sensitive K-Anonymity with Generalization Constraints

Transactions on Data Privacy, Aug 1, 2010

Numerous privacy models based on the k-anonymity property and extending the k-anonymity model hav... more Numerous privacy models based on the k-anonymity property and extending the k-anonymity model have been introduced in the last few years in data privacy research: l-diversity, p-sensitive k-anonymity, (α, k)-anonymity, t-closeness, etc. While differing in their methods and quality of their results, they all focus first on masking the data, and then protecting the quality of the data as a whole. We consider a new approach, where requirements on the amount of distortion allowed on the initial data are imposed in order to preserve its usefulness. Our approach consists of specifying quasiidentifiers' generalization constraints, and achieving p-sensitive k-anonymity within the imposed constraints. We think that limiting the amount of allowed generalization when masking microdata is indispensable for real life datasets and applications. In this paper, the constrained p-sensitive k-anonymity model is introduced and an algorithm for generating constrained p-sensitive k-anonymous microdata is presented. Our experiments have shown that the proposed algorithm is comparable with existing algorithms used for generating p-sensitive k-anonymity with respect to the results' quality, and obviously the obtained masked microdata complies with the generalization constraints as indicated by the user.

Fighting fake news spread in online social networks: Actual trends and future research directions

In this paper we present how fake news spread in the current online social networks. We discuss h... more In this paper we present how fake news spread in the current online social networks. We discuss how existing social network technologies such as influence maximization, information diffusion, and epidemiological models contributes to fake news creation and spreading. Solutions to reducing the creation and spreading of fake news are also reviewed. We make recommendations regarding future areas of research in this field.

On-the-Fly Hierarchies for Numerical Attributes in Data Anonymization

Lecture Notes in Computer Science, 2010

We present in this paper a method for dynamically creating hierarchies for quasi-identifier numer... more We present in this paper a method for dynamically creating hierarchies for quasi-identifier numerical attributes. The resulting hierarchies can be used for generalization in microdata k-anonymization, or for allowing users to define generalization boundaries for constrained k-anonymity. The construction of a new numerical hierarchy for a numerical attribute is performed as a hierarchical agglomerative clustering of that attribute's values in the dataset to anonymize. Therefore, the resulting tree hierarchy reflects well the closeness and clustering tendency of the attribute's values in the dataset. Due to this characteristic of the hierarchies created on-the-fly for quasiidentifier numerical attributes, the quality of the microdata anonymized through generalization based on these hierarchies is well preserved, and the information loss in the anonymization process remains in reasonable bounds, as proved experimentally.

Avoiding Attribute Disclosure with the (Extended) p-Sensitive k-Anonymity Model

Annals of information systems, Oct 15, 2009

Semi-supervised learning techniques: k-means clustering in OODB fragmentation

Vertical and horizontal fragmentation are central issues in the design process of Distributed Obj... more Vertical and horizontal fragmentation are central issues in the design process of Distributed Object Based Systems. A good fragmentation scheme followed by an optimal allocation could greatly enhance performance in such systems, as data transfer between distributed sites is minimized. In this paper we present a horizontal fragmentation approach that uses the k-means AI clustering method for partitioning object instances into fragments. Our new method applies to existing databases, where statistics are already present. We model fragmentation input data in a vector space and give different object similarity measures together with their geometrical interpretations. We provide quality and performance evaluations using a partition evaluator function.

Is Data Collection through Twitter Streaming API Useful for Academic Research?

In this paper we study the reliability of Twitter data collection when Streaming API with filteri... more In this paper we study the reliability of Twitter data collection when Streaming API with filtering is used. For this purpose, we designed a series of experiments that use the free version of Twitter Streaming API. We ran our experiments between June 24 and July 15, 2018, using soccer terms for filtering, due to the popularity of FIFA World Cup competition that was held at that time. Our experiments showed that when filtering is used for terms that are not very popular, then all the matching Tweets are likely provided by Twitter; in this case, analyzing those Tweets will provide reliable results for research purposes. Also, we concluded that concurrent processes that collect filtering Tweets for very popular terms tend to return almost the same Tweets, even if the obtained Tweet set is only a sample of all possible keyword-matching Tweets. We believe that the sampling in the filtering process is deterministic and using the Tweets collected for very popular filtering terms may lead to biased results due to this non-random sampling process.

Preservation of Structural Properties in Anonymized Social Networks

Social networks such as Facebook, LinkedIn, or Twitter have nowadays a global reach that surpasse... more Social networks such as Facebook, LinkedIn, or Twitter have nowadays a global reach that surpassed all previous expectations. Many social networks gather confidential information of their users, and as a result, the privacy in social networks has become a topic of general interest. To defend against privacy violations, several social network anonymization models were introduced. In this paper, we empirically study how well several structural properties of a social network are preserved through an anonymization process. We first anonymize several real and synthetic social networks using the kanonymous cluster social network model, and then we compare how well structural properties such as diameter, centrality measures, clustering coefficients, and topological indices are preserved between the original networks and their anonymized versions. Our experiments show that there are correlations between the structural properties' values obtained from the original network and from the corresponding anonymized networks. Preserving such graph properties through anonymization might be extremely important / essential for subsequent graph-mining of the anonymized networks.

Generating Microdata with P-Sensitive K-Anonymity Property

Springer eBooks, 2007

Data and Structural k-Anonymity in Social Networks

Springer eBooks, 2009

The advent of social network sites in the last years seems to be a trend that will likely continu... more The advent of social network sites in the last years seems to be a trend that will likely continue. What naive technology users may not realize is that the information they provide online is stored and may be used for various purposes. Researchers have pointed out for some time the privacy implications of massive data gathering, and effort has been made to protect the data from unauthorized disclosure. However, the data privacy research has mostly targeted traditional data models such as microdata. Recently, social network data has begun to be analyzed from a specific privacy perspective, one that considers, besides the attribute values that characterize the individual entities in the networks, their relationships with other entities. Our main contributions in this paper are a greedy algorithm for anonymizing a social network and a measure that quantifies the information loss in the anonymization process due to edge generalization.

Community Detection in Anonymized Social Networks

EDBT/ICDT Workshops, 2014

Social media and social networks are embedded in our society to a point that could not have been ... more Social media and social networks are embedded in our society to a point that could not have been imagined only ten years ago. Facebook, LinkedIn, and Twitter are already well known social networks that have a large audience in all age groups. Recently more trendy social sites such as Pinterest, Instagram, Vine, Tumblr, WhatsApp, and Snapchat are being preferred by the younger audience. The amount of data that those social sites gather from their users is continually increasing and this data is very valuable for marketing, research, and various other purposes. At the same time, this data usually contain a significant amount of sensitive information which should be protected against unauthorized disclosure. To protect the privacy of individuals, this data must be anonymized such that the risk of re-identification of specific individuals is very low. In this paper we study how well anonymized social networks preserve existing communities from the original social networks. To anonymize social networks we used two models, namely, k-anonymity for social networks and kdegree anonymity. To determine communities in social networks we used a community detection algorithm based on modularity quality function known as Louvain method. Our experiments on publically available datasets show that anonymized social networks satisfactorily preserve the community structure of their original networks.

Preserving Communities in Anonymized Social Networks

Transactions on Data Privacy, Apr 1, 2015

Social media and social networks are embedded in our society to a point that could not have been ... more Social media and social networks are embedded in our society to a point that could not have been imagined only ten years ago. Facebook, LinkedIn, and Twitter are already well known social networks that have a large audience in all age groups. The amount of data that those social sites gather from their users is continually increasing and this data is very valuable for marketing, research, and various other purposes. At the same time, this data usually contain a significant amount of sensitive information which should be protected against unauthorized disclosure. To protect the privacy of individuals, this data must be anonymized such that the risk of re-identification of specific individuals is very low. In this paper we study if anonymized social networks preserve existing communities from the original social networks. To perform this study, we introduce two approaches to measure the community preservation between the initial network and its anonymized version. In the first approach we simply count how many nodes from the original communities remained in the same community after the processes of anonymization and de-anonymization. In the second approach we consider the community preservation for each node individually. Specifically, for each node, we compare the original and final communities to which the node belongs. To anonymize social networks we use two models, namely, k-anonymity for social networks and k-degree anonymity. To determine communities in social networks we use an existing community detection algorithm based on modularity quality function. Our experiments on publically available datasets show that anonymized social networks satisfactorily preserve the community structure of their original networks.

Efficient Approximation Algorithms for Minimum Dominating Sets in Social Networks

IGI Global eBooks, 2021

Social networks are increasingly becoming an outlet that is more and more powerful in spreading n... more Social networks are increasingly becoming an outlet that is more and more powerful in spreading news and influence individuals. Compared with other traditional media outlets such as newspaper, radio, and television, social networks empower users to spread their ideological message and/or to deliver target advertising very efficiently in terms of both cost and time. In this article, the authors focus on efficiently finding dominating sets in social networks for the classical dominating set problem as well as for two related problems: partial dominating sets and d-hop dominating sets. They will present algorithms for determining efficiently a good approximation for the social network minimum dominating sets for each of the three variants. The authors ran an extensive suite of experiments to test the presented algorithms on several datasets that include real networks made available by the Stanford Network Analysis Project and synthetic networks that follow the power-law and random models that they generated for this work. The performed experiments show that the selection of the algorithm that performs best to determine efficiently the dominating set is dependent of network characteristics and the order of importance between the size of the dominating set and the time required to determine such a set.

A Clustering Approach for Achieving Data Privacy

DMIN, 2007

Search and Retrieval in Semantic-Structural Representations of Novel Malware

Advances in Artificial Intelligence and Machine Learning, 2024

In this study, we present a novel representation for binary programs which captures semantic simi... more In this study, we present a novel representation for binary programs which captures semantic similarity and structural properties. This representation enables the search and retrieval of binary executable programs based on their similarity of behavioral properties. The proposed representation is composed in a bottom-up approach: we begin by extracting data dependency graphs (DDG), which are representative of both program structure and operational semantics. We then encode each program as a set of graph hashes representing isomorphic uniqueness, a method we have labeled DDG Fingerprinting. We present experimental results of search using k-Nearest Neighbors in a metric space constructed from a set of binary executables. Searches in the dataset are based on the operational semantics of specific malware examples. 2052

Sensitive Attributes based Privacy Preserving in Data Mining using k-anonymity

knowledge from huge amount of data. In recent years, there has been a tremendous growth in the am... more knowledge from huge amount of data. In recent years, there has been a tremendous growth in the amount of personal data that can be collected and analyzed by the organizations. Organizations such as credit card companies, real estate companies and hospitals collect and hold large volumes of data for their research purposes. E.g. National Institute of health. When these organizations publish data containing a lot of sensitive information. The importance of sharing data for research and knowledge discovery has been well-recognized. However, sharing data that contains sensitive personal information, such as insurance data, medical record, etc across organization boundaries can raise serious privacy concerns. There is a need to preserve the privacy of the individuals in data set . K-anonymity is one of the easy and efficient techniques to achieve privacy in many data publishing applications. In k-anonymity, all tuples of releasing database are generalized to make it anonymize which lead to data utility reduction and more information loss of publishing table. Sensitive attribute based anonymity method is very useful in preserving the privacy of individuals in organization’s publication of data. It reduces information loss to the researchers by providing sensitive levels. This method also avoids Homogeneity attack and Background attacks.

Efficient Approximation Algorithms for Minimum Dominating Sets in Social Networks

International Journal of Service Science, Management, Engineering, and Technology, Apr 1, 2018

Socialnetworksareincreasinglybecominganoutletthatismoreandmorepowerfulinspreading newsandinfluenc... more Socialnetworksareincreasinglybecominganoutletthatismoreandmorepowerfulinspreading newsandinfluenceindividuals.Comparedwithothertraditionalmediaoutletssuchasnewspaper, radio,andtelevision,socialnetworksempoweruserstospreadtheirideologicalmessageand/orto delivertargetadvertisingveryefficientlyintermsofbothcostandtime.Inthisarticle,theauthors focusonefficientlyfindingdominatingsetsinsocialnetworksfortheclassicaldominatingsetproblem aswellasfortworelatedproblems:partialdominatingsetsandd-hopdominatingsets.Theywill presentalgorithmsfordeterminingefficientlyagoodapproximationforthesocialnetworkminimum dominatingsetsforeachofthethreevariants.Theauthorsrananextensivesuiteofexperimentsto testthepresentedalgorithmsonseveraldatasetsthatincluderealnetworksmadeavailablebythe StanfordNetworkAnalysisProjectandsyntheticnetworksthatfollowthepower-lawandrandom modelsthattheygeneratedforthiswork.Theperformedexperimentsshowthattheselectionofthe algorithmthatperformsbesttodetermineefficientlythedominatingsetisdependentofnetwork characteristics and the order of importance between the size of the dominating set and the time requiredtodeterminesuchaset.

Exploring Sentiment on Campus. A Twitter Sentiment Analysis on University Tweets

EPiC series in computing, Mar 21, 2024

We report in this work the results of our analysis of accuracy of 5 sentiment analysis methods (T... more We report in this work the results of our analysis of accuracy of 5 sentiment analysis methods (TextBlob, VADER, logistic regression, support vector machine, CNN on encodings based on BERT tokenization,) for a dataset consisting of tweets from the academia domain, that we API-scraped for 32 universities during the year 2022. We show some results for the volume and sentiment polarity trends exhibited by this dataset. We connect peak and low sentiment averages to concrete events that explain the respective sentiment trend; this proves that observing the social media trends allows to detect real events that need attention and possible action.

A new approach for optimal fragmentation order in distributed object oriented databases

Pure mathematics and applications, 2004

Class fragmentation is an important task in the design of Distributed OODBs and there are many al... more Class fragmentation is an important task in the design of Distributed OODBs and there are many algorithms handling it. Almost none of them deals however with the class fragmentation order details. We claim that class fragmentation order can induce severe performance penalties if not considered in the fragmentation phase. We propose here two variants of an algorithm for finding the optimal class fragmentation order in a DOODB. In both variants we capture all class relations (inheritance, aggregation) and we determine a class fragmentation order where precedence conflicts induced by relation cycles are eliminated in such way that the strongest links be maintained.

A comparison between local and global recoding algorithms for achieving microdata -sensitive -anonymity

Acta Universitatis Apulensis. Mathematics - Informatics, 2008

New privacy regulations together with ever-increasing data availability and computational power h... more New privacy regulations together with ever-increasing data availability and computational power have created a huge interest in data privacy research. One major research direction is built around k-anonymity property, which is required for the released data. Although k-anonymity protects against identity disclosure, it fails to provide an adequate level of protection with respect to attribute disclosure. We introduced a new privacy protection property called p-sensitive k-anonymity that avoids this shortcoming. We developed new algorithms (GreedyPKClustering and EnhancedPKClustering) and adapted an existing algorithm (Incognito) to generate masked microdata with p-sensitive k-anonymity property. All these algorithms try to reduce the amount of information lost while transforming data to conform to p-sensitive k-anonymity. They are different in the masking methods they use. The new algorithms are based on local recoding masking methods. Incognito, initially designed for k-anonymity, uses global recoding for masking. This paper's goal is to compare the impact of the masking method on the quality of the masked microdata obtained. For this we compare the quality of the results (cost measures based on data utility) and the efficiency (running time) of these three algorithms for masking both real and synthetic data sets.

Efficient approximation algorithms to determine minimum partial dominating sets in social networks

In this paper we report on extensive experiments for determining partial dominating sets of small... more In this paper we report on extensive experiments for determining partial dominating sets of small size for various types of real and synthetic social networks. Our experiments ran on several real network datasets made available by the Stanford Network Analysis Project and on some synthetic power-law and random networks created with social network generators. To compute partial dominating sets on these networks we used five algorithms compared in [4], which were adapted for partial dominating sets. Our experiments showed that there are several good algorithms that can efficiently find quality approximations for the minimum-size partial dominating set problem. The best algorithm choice is dependent on the network characteristics and the value of the coverage parameter.

P-Sensitive K-Anonymity with Generalization Constraints

Transactions on Data Privacy, Aug 1, 2010

Numerous privacy models based on the k-anonymity property and extending the k-anonymity model hav... more Numerous privacy models based on the k-anonymity property and extending the k-anonymity model have been introduced in the last few years in data privacy research: l-diversity, p-sensitive k-anonymity, (α, k)-anonymity, t-closeness, etc. While differing in their methods and quality of their results, they all focus first on masking the data, and then protecting the quality of the data as a whole. We consider a new approach, where requirements on the amount of distortion allowed on the initial data are imposed in order to preserve its usefulness. Our approach consists of specifying quasiidentifiers' generalization constraints, and achieving p-sensitive k-anonymity within the imposed constraints. We think that limiting the amount of allowed generalization when masking microdata is indispensable for real life datasets and applications. In this paper, the constrained p-sensitive k-anonymity model is introduced and an algorithm for generating constrained p-sensitive k-anonymous microdata is presented. Our experiments have shown that the proposed algorithm is comparable with existing algorithms used for generating p-sensitive k-anonymity with respect to the results' quality, and obviously the obtained masked microdata complies with the generalization constraints as indicated by the user.

Fighting fake news spread in online social networks: Actual trends and future research directions

In this paper we present how fake news spread in the current online social networks. We discuss h... more In this paper we present how fake news spread in the current online social networks. We discuss how existing social network technologies such as influence maximization, information diffusion, and epidemiological models contributes to fake news creation and spreading. Solutions to reducing the creation and spreading of fake news are also reviewed. We make recommendations regarding future areas of research in this field.

On-the-Fly Hierarchies for Numerical Attributes in Data Anonymization

Lecture Notes in Computer Science, 2010

We present in this paper a method for dynamically creating hierarchies for quasi-identifier numer... more We present in this paper a method for dynamically creating hierarchies for quasi-identifier numerical attributes. The resulting hierarchies can be used for generalization in microdata k-anonymization, or for allowing users to define generalization boundaries for constrained k-anonymity. The construction of a new numerical hierarchy for a numerical attribute is performed as a hierarchical agglomerative clustering of that attribute's values in the dataset to anonymize. Therefore, the resulting tree hierarchy reflects well the closeness and clustering tendency of the attribute's values in the dataset. Due to this characteristic of the hierarchies created on-the-fly for quasiidentifier numerical attributes, the quality of the microdata anonymized through generalization based on these hierarchies is well preserved, and the information loss in the anonymization process remains in reasonable bounds, as proved experimentally.

Avoiding Attribute Disclosure with the (Extended) p-Sensitive k-Anonymity Model

Annals of information systems, Oct 15, 2009

Semi-supervised learning techniques: k-means clustering in OODB fragmentation

Vertical and horizontal fragmentation are central issues in the design process of Distributed Obj... more Vertical and horizontal fragmentation are central issues in the design process of Distributed Object Based Systems. A good fragmentation scheme followed by an optimal allocation could greatly enhance performance in such systems, as data transfer between distributed sites is minimized. In this paper we present a horizontal fragmentation approach that uses the k-means AI clustering method for partitioning object instances into fragments. Our new method applies to existing databases, where statistics are already present. We model fragmentation input data in a vector space and give different object similarity measures together with their geometrical interpretations. We provide quality and performance evaluations using a partition evaluator function.

Is Data Collection through Twitter Streaming API Useful for Academic Research?

In this paper we study the reliability of Twitter data collection when Streaming API with filteri... more In this paper we study the reliability of Twitter data collection when Streaming API with filtering is used. For this purpose, we designed a series of experiments that use the free version of Twitter Streaming API. We ran our experiments between June 24 and July 15, 2018, using soccer terms for filtering, due to the popularity of FIFA World Cup competition that was held at that time. Our experiments showed that when filtering is used for terms that are not very popular, then all the matching Tweets are likely provided by Twitter; in this case, analyzing those Tweets will provide reliable results for research purposes. Also, we concluded that concurrent processes that collect filtering Tweets for very popular terms tend to return almost the same Tweets, even if the obtained Tweet set is only a sample of all possible keyword-matching Tweets. We believe that the sampling in the filtering process is deterministic and using the Tweets collected for very popular filtering terms may lead to biased results due to this non-random sampling process.

Preservation of Structural Properties in Anonymized Social Networks

Social networks such as Facebook, LinkedIn, or Twitter have nowadays a global reach that surpasse... more Social networks such as Facebook, LinkedIn, or Twitter have nowadays a global reach that surpassed all previous expectations. Many social networks gather confidential information of their users, and as a result, the privacy in social networks has become a topic of general interest. To defend against privacy violations, several social network anonymization models were introduced. In this paper, we empirically study how well several structural properties of a social network are preserved through an anonymization process. We first anonymize several real and synthetic social networks using the kanonymous cluster social network model, and then we compare how well structural properties such as diameter, centrality measures, clustering coefficients, and topological indices are preserved between the original networks and their anonymized versions. Our experiments show that there are correlations between the structural properties' values obtained from the original network and from the corresponding anonymized networks. Preserving such graph properties through anonymization might be extremely important / essential for subsequent graph-mining of the anonymized networks.

Generating Microdata with P-Sensitive K-Anonymity Property

Springer eBooks, 2007

Data and Structural k-Anonymity in Social Networks

Springer eBooks, 2009

The advent of social network sites in the last years seems to be a trend that will likely continu... more The advent of social network sites in the last years seems to be a trend that will likely continue. What naive technology users may not realize is that the information they provide online is stored and may be used for various purposes. Researchers have pointed out for some time the privacy implications of massive data gathering, and effort has been made to protect the data from unauthorized disclosure. However, the data privacy research has mostly targeted traditional data models such as microdata. Recently, social network data has begun to be analyzed from a specific privacy perspective, one that considers, besides the attribute values that characterize the individual entities in the networks, their relationships with other entities. Our main contributions in this paper are a greedy algorithm for anonymizing a social network and a measure that quantifies the information loss in the anonymization process due to edge generalization.

Community Detection in Anonymized Social Networks

EDBT/ICDT Workshops, 2014

Social media and social networks are embedded in our society to a point that could not have been ... more Social media and social networks are embedded in our society to a point that could not have been imagined only ten years ago. Facebook, LinkedIn, and Twitter are already well known social networks that have a large audience in all age groups. Recently more trendy social sites such as Pinterest, Instagram, Vine, Tumblr, WhatsApp, and Snapchat are being preferred by the younger audience. The amount of data that those social sites gather from their users is continually increasing and this data is very valuable for marketing, research, and various other purposes. At the same time, this data usually contain a significant amount of sensitive information which should be protected against unauthorized disclosure. To protect the privacy of individuals, this data must be anonymized such that the risk of re-identification of specific individuals is very low. In this paper we study how well anonymized social networks preserve existing communities from the original social networks. To anonymize social networks we used two models, namely, k-anonymity for social networks and kdegree anonymity. To determine communities in social networks we used a community detection algorithm based on modularity quality function known as Louvain method. Our experiments on publically available datasets show that anonymized social networks satisfactorily preserve the community structure of their original networks.

Preserving Communities in Anonymized Social Networks

Transactions on Data Privacy, Apr 1, 2015

Social media and social networks are embedded in our society to a point that could not have been ... more Social media and social networks are embedded in our society to a point that could not have been imagined only ten years ago. Facebook, LinkedIn, and Twitter are already well known social networks that have a large audience in all age groups. The amount of data that those social sites gather from their users is continually increasing and this data is very valuable for marketing, research, and various other purposes. At the same time, this data usually contain a significant amount of sensitive information which should be protected against unauthorized disclosure. To protect the privacy of individuals, this data must be anonymized such that the risk of re-identification of specific individuals is very low. In this paper we study if anonymized social networks preserve existing communities from the original social networks. To perform this study, we introduce two approaches to measure the community preservation between the initial network and its anonymized version. In the first approach we simply count how many nodes from the original communities remained in the same community after the processes of anonymization and de-anonymization. In the second approach we consider the community preservation for each node individually. Specifically, for each node, we compare the original and final communities to which the node belongs. To anonymize social networks we use two models, namely, k-anonymity for social networks and k-degree anonymity. To determine communities in social networks we use an existing community detection algorithm based on modularity quality function. Our experiments on publically available datasets show that anonymized social networks satisfactorily preserve the community structure of their original networks.

Efficient Approximation Algorithms for Minimum Dominating Sets in Social Networks

IGI Global eBooks, 2021

Social networks are increasingly becoming an outlet that is more and more powerful in spreading n... more Social networks are increasingly becoming an outlet that is more and more powerful in spreading news and influence individuals. Compared with other traditional media outlets such as newspaper, radio, and television, social networks empower users to spread their ideological message and/or to deliver target advertising very efficiently in terms of both cost and time. In this article, the authors focus on efficiently finding dominating sets in social networks for the classical dominating set problem as well as for two related problems: partial dominating sets and d-hop dominating sets. They will present algorithms for determining efficiently a good approximation for the social network minimum dominating sets for each of the three variants. The authors ran an extensive suite of experiments to test the presented algorithms on several datasets that include real networks made available by the Stanford Network Analysis Project and synthetic networks that follow the power-law and random models that they generated for this work. The performed experiments show that the selection of the algorithm that performs best to determine efficiently the dominating set is dependent of network characteristics and the order of importance between the size of the dominating set and the time required to determine such a set.

A Clustering Approach for Achieving Data Privacy

DMIN, 2007