Business Intelligence meets Big Data: An Overview on Security and Privacy

Size: px

Start display at page:

Download "Business Intelligence meets Big Data: An Overview on Security and Privacy"

Irene Stewart
8 years ago
Views:

1 Business Intelligence meets Big Data: An Overview on Security and Privacy Claudio A. Ardagna Ernesto Damiani Dipartimento di Informatica - Università degli Studi di Milano NSF Workshop on Big Data Security and Privacy September 16-17, 2014 Dallas, TX, USA 1/15

Ardagna Ernesto Damiani Dipartimento di Informatica - Università

2 Motivation (1) Every minute, the world generates 1.7 million billion bytes of data, that is, over 6 megabytes for each human being on the planet Earth Widespread diffusion of Cloud Computing and Internet of Things Users increasingly access remote services and is surrounded by sensors of different types Big Data is a trend topic: studied in several domains (research, business, military, government), discussed in TV shows, and mentioned in TV Series The business potential of big data is huge 2/15

Cloud Computing and Internet of Things Users increasingly access remote services and is surrounded by sensors of different

3 Motivation (2) The European Commission s recent Communication on the data-driven economy describes its new strategy to promote the data-driven economy in the European Union Technical improvements of computational and storage technologies support new business intelligence Business domains are likely to be reshaped by analytics Many organizations worldwide already collect and analyze their own business process data to improve their internal decision making Few organizations worldwide are able to collect and analyze big data 3/15

intelligence Business domains are likely to be reshaped by analytics Many organizations worldwide already collect and analyze their

4 Next-generation business intelligence (1) Increasing demand both by research and industry communities to Move traditional business intelligence (data warehousing, latency, data sampling) to big data Identify practical key scientific and technical challenges to enable big data business intelligence Exploit big data to increase security, reliability, and safety of a distributed system Guarantee privacy of big data from unauthorized use and unwanted inference 4/15

scientific and technical challenges to enable big data business intelligence Exploit big data to increase security,

5 Next-generation business intelligence (2) Twin notions of full data and zero latency are value propositions underlying a new notion of business intelligence Full data refers to the availability of all relevant information Zero latency refers to the support for real-time analysis of data Zero-latency analysis of full data means better decisions and more accurate predictions A competitive gap between the haves, who have fast access to huge amounts of data, and the have-nots, who have to live with high latency and small data samples 5/15

real-time analysis of data Zero-latency analysis of full data means better decisions and more accurate predictions A competitive gap

6 Challenges (1) Manifesto on Business Intelligence Meets Big Data defines 9 challenges for accurate decision making and prediction techniques Challenge 1: Data preparation, quality, and trustworthiness [S,P] Availability of high-quality, precise, and trustworthy data Techniques for their management to increase the added value of information extracted by big data analytics techniques Challenge 2: Efficient distributed storage and search [ ] Timeliness of data collection for prompt analysis of big data Need to reduce potential sources of latency and enhance search algorithms Challenge 3: Effective online data analysis (BIG-OLAP) [P] Online analysis of multidimensional data Adapt existing OLAP approaches to big data 6/15

data analytics techniques Challenge 2: Efficient distributed storage and search [ ] Timeliness of data collection for prompt analysis of big data Need to reduce potential sources of

7 Challenges (2) Manifesto on Business Intelligence Meets Big Data defines 9 challenges for accurate decision making and prediction techniques Challenge 4: Effective machine learning techniques for big data mining [S,P] Machine learning and data mining to unleash the full potential of collected information Challenge 5: Efficient handling of big data streams [S,P] Support inference on data streams coming from different types of sensors/devices Challenge 6: Semantic lifting techniques [P] Cope with semantics of big data for fine-grained inference 7/15

full potential of collected information Challenge 5: Efficient handling of big data streams [S,P] Support inference on data streams coming

8 Challenges (3) Manifesto on Business Intelligence Meets Big Data defines 9 challenges for accurate decision making and prediction techniques Challenge 7: Programming models [S] Different models supporting management, reliability, scalability of big data infrastructures MapReduce, Apache Hadoop, and many others Challenge 8: Social analytics [P] Widespread diffusion of social media and availability of social-related data Analytics solutions for modeling and filtering social interaction data Challenge 9: Security and privacy [S,P] More in the next slides... 8/15

Apache Hadoop, and many others Challenge 8: Social analytics [P] Widespread diffusion of social media and availability of social-related data

9 Security and privacy in big data (1) Security Big data provide priceless sources of information at the basis of robust and accurate security solutions Privacy Big data contain sensitive information that needs to be protected from unauthorized access and release Proper solutions should find a balance between the needs of security and privacy in a big data scenario Full data and zero latency fundamental to increase security providing better decisions and more accurate predictions Full data conflicts with privacy protection 9/15

Proper solutions should find a balance between the needs of security and privacy in a big data scenario Full data and zero latency

10 Security and privacy in big data (2) The more the data available for inference, the more the quality and precision of security techniques and, in turn, the security of the system implementing them With full data availability, system security should achieve the optimum level The overhead given by data communication prior to analysis could invalidate the advantages given by the availability of full data Privacy can be compromised by the availability of full data 10/15

security should achieve the optimum level The overhead given by data communication prior to analysis could invalidate

11 Use case 1: Attack detection and classification (1) Vulnerability assessment and cyber security Non real-time sharing of cyber security information is the state-of-the-art New approaches for real time-sharing of cyber security data amongst (more or less trusted) collaborating organizations can increase security Shared information can enable the detection of events in a collaborative effort Need of a solution with fast and accurate algorithms for big data analysis and with access to the wider amount of data possible 11/15

trusted) collaborating organizations can increase security Shared information can enable the detection of events in a collaborative

12 Use case 1: Attack detection and classification (2) Main challenges Provide analysis methods performing the detection of attacks and anomalies based on the shared data Security information must be ready for online and real-time decisions Perform analytics on full data (strictly related to the need of zero latency solutions) Zero latency difficult to achieve when big data are involved Protect privacy of data by applying proper filtering, anonymization, sanitization, or pseudonymization methods Long-Term expectations SMEs and individuals could benefit from a significantly increased cyber protection level The more the information is fresh and up-to-date, the more the implemented defenses are effective and zero-day exploits less probable 12/15

big data are involved Protect privacy of data by applying proper filtering, anonymization, sanitization, or pseudonymization methods Long-Term expectations SMEs and individuals could

13 Use case 2: Online auctioning (1) An auctioneer a 1 selling a good g at price p A buyer b looking for g and willing to pay p >p for it a 1 b offer p to buy g OK Pay p for g OK,g 13/15

14 Use case 2: Online auctioning (2) Zero latency can also be exploited by malicious attackers An auctioneer a 2 working at zero latency 1. Offers p, such that p >p >p, to a 1 for g 2. Buys g 3. Sells g to b for p and gains p p a 1 a 2 b offer p to buy g OK offer p to buy g Pay p for g OK OK,g Pay p for g OK,g 14/15

Offers p, such that p >p >p, to a 1 for g 2. Buys g 3.

15 Conclusions Security and privacy Are among the most important requirements in a big data scenario Are conflicting and need to be carefully balanced to find an optimum approach Security can take advantage by an increasing amount of available data that can be analyzed in real time Privacy can be highly affected and suggests users to limit the free and unregulated sharing of their data Next-generation business intelligence approaches for big data should Revisit existing security and privacy approaches Find the best compromise between the need of security and privacy protection, and the overhead posed by techniques for the analysis of big data Support assurance and certification techniques 15/15

the free and unregulated sharing of their data Next-generation business intelligence approaches for big data should Revisit existing security and privacy approaches Find the

Business Intelligence meets Big Data: An Overview on Security and Privacy

Business Intelligence meets Big Data: An Overview on Security and Privacy Claudio A. Ardagna and Ernesto Damiani Università degli Studi di Milano, 26013 Crema, Italy {firstname.lastname}@unimi.it Abstract.