It’s Time To Get Serious About Big Data Security

August 27, 2013
No Comment


Over the last few years, we have witnessed a substantial amount of buzz about Big Data, and we have seen the large-scale adoption of Big Data analytics solutions. Organizations are collecting and analyzing massive amounts of data in a way that most were unable to do ten years ago. Pattern recognition, non-obvious relationship detection, graph analysis, predictive processing, and other analytical approaches can enable our organizations to turn oceans of data into knowledge, allowing us to make the best use of the data that we have.

At the same time, there exists an ever-growing concern that revolves around the security and protection of the information that we are analyzing. As we watch the explosive growth of data that we are processing from multiple disparate data sources, our IT environments are growing increasingly more complex, and we are witnessing new challenges related to data management, security, and privacy. Many organizations are now required to enforce access control and privacy restrictions in order to meet data and privacy regulations, and these same organizations face steep penalties and fines for non-compliance. At the same time, data security breaches (both from the insider and outsider) are on the rise, and a study released this year by Symantec and the Ponemon institute found that the average organizational cost of one security breach in the United States is 5.4 million dollars. Civil court records show a growing list of organizations that have faced millions of dollars in lawsuits related to their exposure of information, and new studies are showing that most organizations are not prepared to respond to the fallout.

So what does that mean for organizations focusing on Big Data? With evolving cyber security threats, growing security and privacy concerns, the reality of insider attacks, and the legal and regulatory requirements that many organizations face, it means that we need to start getting serious about Big Data Security. It’s an issue – but one that many organizations haven’t thought about.

Although security and privacy issues related to Big Data are starting to get more attention in the mainstream media, the issues are not new – many of these concerns have been documented in academic papers in the Data Mining and Knowledge Discovery communities for decades. As distributed analytics technologies have become more prevalent and available in recent years, these concerns that were once raised in academia are now being realized in Big Data implementations.

Further complicating the matter is that many Big Data platforms, like Apache Hadoop, were not originally designed with security in mind. Many security professionals over the years have highlighted challenges related to Hadoop’s security model, and as a result there has been an explosive growth in security-focused tools that complement Hadoop offerings, with products like Cloudera SentryIBM InfoSphere Optim Data MaskingIntel’s secure Hadoop distributionDataStax EnterpriseDataGuise for HadoopProtegrity Big Data Protector for HadoopRevelytix LoomZettaset Secure Data Warehouse — and the list could go on. At the same time, Hadoop’s security model is slowly evolving to meet some of these security challenges because of some great work happening in the source community.

Regardless of your Big Data platform, however, there are some critical steps that organizations must do in order to understand and be prepared for Big Data security challenges.

  1. Identify and understand the sensitivity levels of your data. In order to understand the challenges, it is critical to understand the sensitivity levels of your data and the access control policies associated with your data. The sensitivity levels of your information will drive your security strategy. In some cases, it may be necessary to filter access to the data retrieved by data scientists and analysts running jobs and queries, based on what they are allowed to see. As data sets from different information sources are combined with others for processing – where each data set may have inherent access control policies – defining and enforcing the access control policies on the combined sets can become a tricky problem.
  2. Understand the impact of the release of your data. Even organizations that are careful to strip out sensitive information from data sets that they release to their business partners are finding that their data sets are vulnerable to attacks related to differential privacy, and many companies have learned this the hard way. A few years ago, for example, Netflix anonymized movie ratings of about 480,000 customers, and held a contest, offering 1 million dollars for the contestant who could improve their “Recommended for You” movie suggestions. A few researchers found that by combining anonymous ratings in the Netflix data set with public ratings in the IMDB data set, they could successfully de-anonymize the ratings of many viewers, exposing information about the sexual interests, political leanings, and religious views of many viewers. The publication of their findings led to a multi-million dollar lawsuit against Netflix, which Netflix later settled.
  3. Develop policies & procedures related to Big Data Security and Privacy. These will vary from organization to organization, based on the sensitivity levels and inherent release policies of your data, but should include policies and procedures related to data ingest and release, access control within the organization, data destruction and sanitation, monitoring procedures, and policies and procedures for incident response.   
  4. Develop and execute a technical security approach that complements the security of your analytics platforms.  Security controls on many Big Data platforms aren’t providing the amount of security that some of our organizations need.  Requirements for authorization at the data level, on-disk encryption, and integration with Identity and Access Management infrastructure, as well as  proactive monitoring of your data (just to name a few) often need to be solved by tools that complement your analytics platform. Many times, certain requirements may involve the integration of third party encryption solutions into your analytics platforms or the use of other tools built on top of them to satisfy other requirements (Apache Accumulo provides cell-based authorization over top of Hadoop, for example). Because of concerns related to security, some organizations segregate their analytics clusters on their internal networks, and provide perimeter access control to authorized users. While this certainly provides a certain level of protection, it doesn’t provide authorization at the data level mandated by some organizations. Because of the complexity of distributed analytics platforms, security configuration is often tricky and complex.  Developing security solutions for analytics platforms therefore requires much attention to detail, a certain degree of creativity, and requires that you develop a security approach that revolves around the security policies of your organization and data.

These steps towards “getting serious about Big Data Security” are extremely necessary in our interconnected digital world, where the cost of data breaches are rising, and the penalties of not protecting our information are steep –  they certainly affect our organization’s budget outlook, but more importantly, security breaches will affect our reputation.  As security and privacy concerns around Big Data continue to attract attention, look for much innovation to occur in this exciting technology space.

About the Author

Kevin T. Smith is the Director of Technology Solutions and Outreach for the Applied Mission Solutions division of Novetta Solutions, where he provides strategic technology leadership and develops innovative, data-focused and highly-secure solutions for customers. He is the author of numerous technology articles, including a recent article at InfoQ : Big Data Security – The Evolution of Hadoop’s Security Model. He has authored many technology books, including the upcoming book Professional Hadoop Solutions (Wrox Press) as well as Applied SOA: Service-Oriented Architecture and Design Strategies (Wiley), The Semantic Web: A Guide to the Future of XML, Web Services, and Knowledge Management (Wiley), and many others. He can be reached at

Via CTO Vision