Center for Open Innovation :Â With focus on Smart Cities and entrepreneurship
Center for Open Innovation is launched at AU Campus by Hon’ble CM Sri Chandrababu Naidu on 10th January, 2016.
Garwood Centre for Corporate Innovation, Haas School of Business, University of California, Berkley and the proposed Centre of Open Innovation, Andhra University would provide an ecosystem for collaborating with knowledge flows globally.
MIT (Boston) 02-11-2015
An Expression of Interest was entered among Andhra University, AP Innovation Society and MIT spin off FAB Lab to set up SUPER FAB LAB in Andhra University, Visakhapatnam – first of its its kind of Super Fab Lab in South India – to promote design incubation among young students and innovators.
MIT is ranked as best institution for engineering and technology in the world. It is a great moment for AP government to bring MIT spin off to the State of Andhra Pradesh.
Excited to have this lab in Andhra University very soon 🙂
For more details download the USA Tour Report Prepared by AP Innovation Society. Courtesy: AP IT & EC web site
A few photos of the trip can also be found here:
Posted by Palle Raghunath Reddy on Monday, November 2, 2015
IMPROVING UTILITY IN PRIVACY PRESERVING PUBLISHING OF RELATIONAL AND SOCIAL NETWORK DATA
Adusumilli Sri Krishna
The digitization of our everyday lives has led to an explosion in the collection of individual data by various public, private and non-profit organizations. These organizations publish the data over internet to extend the accessibility to the out- side world, research and public for the purpose of data analysis. Privacy is a double edged sword, there should be adequate privacy to make sure that sensitive information about the individuals is not disclosed by the published data and at the same time there should be meaningful data to perform the data analysis.
With more attention being given to privacy, researchers are attracted to de- sign new privacy models and anonymization algorithms for privacy preserving data publication. Besides accomplishing the privacy guarantee as defined by the privacy model, data utility is another basic requirement that any anonymization method should satisfy. It is noteworthy to ensure that the changes done at anonymization process does not influence the quality of the data analysis. In spite of several efforts in current research works still many outstanding issues remain to be solved.
This thesis aims to contribute a few techniques for data anonymization while optimising the privacy-utility tradeoff. Specifically, the proposed models deal with emerging privacy issues in relational data and social network data. These two types of data are derived from activities of our everyday life, thus the privacy of such data is important. The main objective of this thesis work is divided into three research problems.
First research problem, addresses the issues involved in creating concept hier- archy trees. Concept hierarchy trees are models created with the help of domain experts and are used in data anonymization. The manual construction of concept hierarchy trees might cause problems such as erroneous classifications or omissions of concepts. This research proposes an approach to construct the concept hierarchy tree dynamically by considering the normalized distance between the nodes and frequency of the attribute values. This method produces better concept hierarchies for both generalization and suppression for anonymization process which significantly improved the utility of the anonymized data when compared to existing methods.
Performance evaluation is done for the proposed method Dynamic Concept Hierarchy Tree (DCHT) against known methods, k -member clustering anonymization and Mondrian multi-dimensional algorithm using 1) Improved on the fly hierarchy (IOTF) 2) On the fly hierarchy (OTF) 3) Hierarchy free (HF) 4) Predefined hierarchy (PH) 5) CHU 6) HAN methods. The metrics used for evaluation are a) Information loss, b) Discernibility metric, c) Normalized average equivalence size metric. Experimental results indicate that DCHT approach is more effective and flexible and the utility is 12% better than IOTF, 16% better than OTF & CHU, 17% better than PH and 21% better than HAN methods when applied on Mondrain multi-dimensional algorithm.
Second research problem addresses the problems encountered in the process of k-anonymization. Most of these were found to be performing over-generalization because full-domain and sub-tree generalization techniques were used. This thesis overcomes over-generalization by defining priority for each quasi-identifer attribute based on concept hierarchy tree and their value distribution in the given data. It balances the privacy and utility tradeoff in data anonymization process by em- ploying cell level generalization or local recoding scheme.
The proposed Priority Based Local Recoding (PBLR) algorithm gives the lowest information loss for all k values and always produces equivalence classes with sizes very close to the specified k compared to Incognito algorithm. The information loss of the PBLR algorithm was found to be 3.73 times lower than that of the Incognito algorithm.
Third research problem, focuses on sensitive attribute disclosure in social net- work data publication. The aim is to propose a new privacy model that provides adequate privacy protection and retains enough data utility in social network data publication. Specifically, this proposed (α,k)-Safe anonymity model, protects both identity and sensitive attribute disclosure in the context of different background knowledges of an adversary. To achieve this (α, k)-Safe anonymity Sensitive Safe Anonymization (SESAAN) algorithm is proposed for large scale graph data.
The effectiveness of SESAAN is verified by conducting extensive set of experiment on real data. The performance of SESAAN is measured in terms of number of edges added or deleted for different k-values with α=0.5. The anonymized graph utility is measured in terms of degree distribution, clustering coefficients and average path length.
Cooperative Game Theory Based Privacy Preserving Techniques for Data Publishing
L. S Chakravarty
Awarded 2015
In recent years, the nature of the adversary’s behaviour is a witness for the drastic increase of the need for new mechanisms demanding Privacy Preserving Data Publishing (PPDP). Meeting this demand is quite challenging since privacy is subject to several constraints such as adversary background knowledge, information loss, and utility. For instance, in record linkage problem of privacy the adversary can easily identify the individual’s sensitive information like age, disease, by linking
the published data with other data sources.
Enormous data are being collected on different aspects of people’s lives such as their credit card transaction records, phone call and email lists, personal health information and web browsing habits. Most of this data is to be scanned for important information for other purposes like evolution of credit risk. Such analysis of private information often raises concerns regarding the privacy rights of individuals’ and organizations. Disobedience to privacy protocols and adversary’s background information may lead to privacy attacks. In this regard, privacy community people have proposed different noteworthy principles like k-Anonymity, l-Diversity and -Differential privacy.
Cooperation among objects like people, organization etc., has recently emerged as a novel paradigm that can yield tremendous performance gains and enables efficient and new models for privacy preserving data publishing. This thesis focuses on how Cooperative Game Theory (CGT) helps in the study of how an agent behaves, in cooperative manner, in formation of groups in data privacy context.
To overcome some of the shortcomings in k-Anonymity, this work presents new approaches to achieve k-Anonymity using Concept Hierarchy Trees (CHT) based on CGT. The observed shortcomings in existing works are: i) The anonymization is done based on an assumed value of k, ii) the information loss can be found only after anonymization is done, iii) if the information loss is found to be more than the affordable loss then another value of k is to be considered and the whole process of anonymity has to be repeated. To achieve anonymity, coalitions are established between the tuples based on their payoffs which are assigned using (CHT) of
Quasi Identifiers (QID). The new system allows the user to control the required information loss as anonymization metric in data publishing. It also presents a new approach to reduce information loss and in turn reduces anonymization level while achieving k-Anonymity using CHT with weighted levels. The experimental results showing the practicality and scalability are presented.
Co-Operative Privacy Game (CoPG) is proposed in this work to achieve data
privacy. The main objective of the CoPG is to obtain individual’s (player) privacy as a goal that is rationally interested in other individuals (players) privacy. CoPG is formally defined in terms of Nash equilibria, i.e., all the players are in their best coalition, to achieve k-anonymity. The Cooperative Value (CoV ) of each tuple are measured using the characteristic function of the CoPG to identify the coalitions. As the underlying game is convex; the algorithm is efficient and yields high quality coalition formation with respect to intensity and disperse. The variations of the information loss with the parameters β (weight factor of ’nearness’) and γ (multiplicity) are analyzed and the obtained results are discussed.
Attended the Gate closing ceremony, Wagah Border, 9th August, 2015
I used to watch along with my dad the beating retreat on television. Later, too used to watch frequently on youtube like the one here.
This trip gave me an opportunity to watch it in live. I became a school going kid once I saw the flag and the vibrant atmosphere here. I really could not stop myself running on path leading to the gate shouting aloud ‘JAI HIND’
And one of the senior persons accompanying us was all smiles and took this picture and later whatsapped me 🙂
Really the true feeling of Nation comes here.
