Data Infrastructure for Next-Gen Finance

Stage 1: Proof of Concept (High Vulnerability)

Most organizations begin with a proof of concept. At this stage, only a few people will be working with a single dataset; these are people who are trying to understand the performance and insights Hadoop is capable of providing. In this scenario, anyone who has access to the cluster has access to all of the data, but the number of users is quite limited. “At this point, security is often just a firewall, network segregation or what’s called ‘air gapping’ the cluster,” said Heywood.

Stage 2: Live Data with Real Users (Ensuring Basic Security Controls)

The next stage involves live datasets. This is when you need to ensure that you have the security controls in place for strong authentication, authorization, and auditing. According to Heywood, this security will control who can log into the cluster, what the dataset will include, and provide an understanding of everything else that is going on with the data.

Stage 3: Multiple Workloads (Data Is Managed, Secure, and Protected)

Now you’re ready to run multiple workloads. At this stage, you are running a multitenant operation and therefore you need to lock down access to the data; authorization and security controls are even more important at this juncture.

The goal at this stage, explained Heywood, is to be able to have all of the data in a single enterprise data hub and allow access to the appropriate audiences, while simultaneously ensuring that there are no incidental access issues for people who shouldn’t have access to given datasets. Additionally, you will need to provide a comprehensive audit.

Stage 4: In Production at Scale (Fully Compliance Ready)

After you have a production use case in place, you can begin moving over other datasets. You are finally ready to run at scale and in production with sensitive datasets that might fall under different types of regulations. According to Heywood, this is when you need to run a fully compliance-ready stack, which includes the following:

• Encryption and key management in place

• Full separation of duties

• Separate sets of administrators to configure the parameter, control the authorization layer, and conduct an ongoing audit

• A separate class of users who are managing the keys tied to the encryption

That’s an overview of what it takes to run a fully compliant data hub. Now let’s take a look at the tools to get there.

Cloudera Manager (Authentication)

Using Cloudera Manager, you can configure all the Kerberos authentication within your environment. However, very few people know how to configure and deploy a Kerberos cluster. As a result, MasterCard automated the configuration process, burying it behind a point-and-click interface.

Apache Sentry (Access Permissions)

After you have authenticated the user, how do you control what they have access to? This is where Apache Sentry comes in. According to Heywood, Apache Sentry provides role-based access control that’s uniformly enforced across all of the Hadoop access paths. You specify your policies and the policies grant a given set of permissions. Then, you link Active Directory groups to those Sentry roles or policies, and users in those groups get access to those datasets.

According to Heywood, what makes this method powerful is that the roles provide an abstraction for managing the permissions. When multiple groups need identical sets of access within a cluster, you don’t need to configure the permissions for each of those groups independently. “Instead, you can set the access policy and relate all those groups back to that one policy; this creates a much more scalable, maintainable way to manage permissions,” Heywood said.

Cloudera Navigator (Visibility)

“Even if you have strong authentication and authorization in place, you need to understand what’s happening within the cluster,” said Heywood. For example:

• What are users doing?

• Who’s accessing a given asset?

• How are those assets being copied around?

• Where is sensitive data being copied to in new files?

This falls under the visibility pillar. Cloudera Navigator provides a comprehensive audit of all the activities that are taking place within the cluster and is the tool of choice for MasterCard.

HDFS Encryption (Protection)

The compliant data protection is encryption. “With HDFS encryption, you can encrypt every single thing within the cluster for HDFS and HBase,” said Heywood. This encryption is end-to-end, meaning that data is encrypted both at-rest and in-flight; encryption and decryption can be done only by the client. “HDFS and HDFS administrators never have access to sensitive key material or unencrypted plain text, further enhancing security,” said Heywood.

Cloudera RecordService (Synchronization)

Cloudera has recently introduced RecordService, a tool that delivers fine-grained access control to tools such as Spark. According to Heywood, “If you had a single dataset and you had to support multiple audiences with different views on that dataset through MapReduce or Spark, typically what you had to do is create multiple copies of the files that were simply limited to what the given audience should have, and then give the audience access to that one file.” This would normally be very difficult primarily due to synchronization issues and keeping copies up to date. RecordService eliminates those problems. “You configure the policy in Sentry, it’s enforced with RecordService, and no matter which way the user is trying to get to the data, we will uniformly and consistently provide the level of access control that’s appropriate for the policy,” said Heywood.

Looking for Lineage

MasterCard knew that lineage and an audit trail were imperatives. “We are about security in everything we do,” said Curcuru. “From our CEO who tells us that we are stewards of people’s digital personas, all the way down to individuals who are actually accessing that data.”

MasterCard’s tools of choice? Cloudera Navigator, which is native in Hadoop, Teradata, Oracle, and SAS. The company advises using native tools first and then looking outside for additional support. After you have the tools in place, what do you need to do to become PCI certified. “You have to create a repeatable process in three areas: people, process, and technology,” said Curcuru.

The following sections describe the areas where MasterCard found people needed the most training.

Segregation of Duties

Segregation of duties matters when you’re conducting an audit or securing a cluster. If there is a breach, you can identify whether it happened by someone with a valid ID. “A lot of the breaches are happening now with valid credentials,” said Curcuru. Gone are the days when one database administrator can have access to your entire system. Similarly, he explained, if your analysts only have access to a certain piece of data and you have a breach, you can isolate where that breach happened. Auditors, he emphasized, will ask about segregation of duties and who can access what.

Documentation

Documentation is critical. Like it or not, if there is a breach, auditors and attorneys are all going to be asking for it. “This is where automated lineage comes into play,” said Curcuru. “You don’t want to be creating documentation manually.”

Awareness Training

Continual knowledge transfer and awareness training are also absolute necessities. For example, warns Curcuru, be aware of the breach that memory sticks create. “Don’t leave those sticks in your hotel room. That type of awareness training, that type of knowledge transfer, has to exist,” he said.

Strong Authentication

Curcuru advocates multifactor authentication, which is a security system that requires more than one method of authentication from independent categories of credentials to verify the user’s identity for a login or other transaction. “It’s not just a domain password and an ID,” said Curcuru. “You’ve got to have something that maybe only you know or the user only knows along with that ID and password. And it can be tokenized.”

Security Logging and Alerts

Hadoop provides the ability to do security logging in real time, which lets you track the behaviors of people who are accessing your data. If analysts do something out of the norm, such as try to get into databases they don’t have access to, you can create proactive alerts in real time.

Continuous Penetration Testing

MasterCard conducts continuous penetration testing, which involves testing a system, network, or application to find vulnerabilities that hackers might exploit. And, according to Curcuru, it’s definitely something that everyone from auditors and your legal team to internal operations will ask about.

Native Data Encryption

Curcuru encourages organizations to look at how they’re handling native data encryption because not all encryption is the same, nor is how you use it. “How the business wants to be able to use the data makes the use of encryption different, as well,” he said.

Embedding Security in Metadata

Embedding security in metadata is particularly important because it makes it possible for the security to follow the metadata through the cluster. Curcuru compares it to taking his daughter to Disney World. “She is with me. I don’t just don’t let her go run willy-nilly.” Similarly, embedding security in metadata lets you to see if the data is crossing multiple machines, when it’s in motion or at rest, and whether there are multiple copies of it. Embedded security in metadata also makes it easier to explain to an auditor what’s going on with your data.

Key Management

Do not store keys with the data. “Key management is important, as well as access to those keys,” said Curcuru. “If you have access to the data, you should not have access to the keys.” Even you are an organization of two or three people, those two roles and responsibilities should be separated.

Keep a Separate Lake of Anonymized Data

MasterCard keeps a data lake separate from a discovery area, and restricts access to it. Data is anonymized and restricted by using role-based access, even if it’s in the sandbox for exploration. For example, an analyst might know that a specific amount of money was spent at a specific store, but she does not know who spent that money.

Ethics

According to Curcuru, the future of data governance involves consent. “It is the stewardship of how the data is used,” he said. “Data governance used to always be in IT. It used to be ETL and taxonomy, but the questions being asked in boardrooms these days involve ethics, i.e., ‘We have this data, but should we use it?’” These questions are being asked not just due to liability issues, but also because of damage to the brand that can be caused by the misuse of data.

Does data governance need to move from descriptive to prescriptive? Apparently the answer is yes, no, and it depends. Curcuru emphasized that the answer boils down to the question of how one wants to use the data. In other words, how does what you’re doing with the data relate back to the business use case, and how was that data originally intended to be used?

Harden agreed that certain controls need to be in place “so that we protect ourselves from ourselves.” But, he added, “We also have a set of tools and processes and data available to us that we’ve never had before. We want to be able to innovate with that and use that data to solve new problems.”

The underlying question that Curcuru says he comes back to is “How do I make sure that I can protect and be stewards of that person who is giving me the privilege of handling their data?” Cunningham agrees that big data governances need to be prescriptive as well as flexible.

Donsky suggested that it’s possible to set up prescriptive governance up front, but that governance needs to allow an element of agility to enable users to do whatever they want within safe environments while also making sure that the enterprise is honoring its customers’ trust.

“The real benefit of Hadoop,” says Donksy, “is that you can store petabytes worth of data in the original fidelity.” A lot of the data wrangling, data preparation, and data consumption is done with more agility in Hadoop, and the net result, he says, is the foundation of governance. He advises, however, that it’s imperative to think about governance at the beginning of a project rather than trying to inject it after the fact.

Machine Learning

Harden suggests that data governance might soon incorporate machine learning techniques. Imagine, he said, being able to take all the information about how data is used in Hadoop, publishing a model of how people are consuming the data, and then building governance capabilities based on that. From there, the system can learn to automatically understand what kind of data is coming into your cluster.

Automated big data governance is certainly the goal, said Harden. The answer of whether it is possible depends on whether anyone is willing to make the initial investments. There are a number of players and tools that are moving in that direction, he says, including Alation, Collibra, FINRA’s open source governance tool Herd, and LinkedIn’s open source data discovery and lineage portal, WhereHows.

According to Harden, Alation combines machine learning with human insight to automatically capture information about what the data describes, where the data comes from, who’s using it, and how it’s used. Collibra focuses on automating data management processes for data stewards, managers, and users. And FINRA’s Herd provides a unified data catalog that helps separate compute from storage in the cloud. Harden explained that Herd also helps track lineage, manage clusters, and automate processing jobs. Herd tracks and catalogs data in a data repository accessible via web service APIs. The repository captures audit and data lineage information to fulfill the requirements of data-driven and highly regulated business environments. LinkedIn’s WhereHows is a frontend system for capturing data on ingest. According to Harden, it provides a central repository and portal that integrates with data processing environments to extract coarse and fine-grained metadata.

Data Quality Management

Cunningham envisions a world where good data quality management and metadata management are incorporated into enterprise operations. According to Harden, “Quality is a component of governance. Having a program in place will drive higher data quality.”

Data Access

Cunningham and Donsky further envision a future in which there are tools that enable and support making data accessible and easier to understand, providing a broader view of serving the whole class of users on top of the foundation that is data governance. Harden adds that “The tools are only as good as the people and processes in place.”

360-Degree Customer View versus Customer Event Hub

In the 360-degree customer view, a customer is surrounded by an ever-increasing set of channels of interaction; the view is an augmentation of all of these channels, all the data, all interactions that are happening with one particular customer across all these different channels. The 360 view brings the data together to create a single view for consumption.

The purpose and advantage of having a 360 view, explained Prabhakar, is that it gives you a consistent understanding of the customer and helps you build relevant functionality. The problem is that these various channels are often implemented as silos and therefore they are isolated from one another, which creates a fragmented user experience. The CEH collapses all these channels into a single omnichannel.

“The key difference between a CEH and a 360-degree customer view is the interactivity,” said Prabhakar. “A 360-degree view is for consumption by the enterprise, whereas a CEH is a bi-directional channel” that allows for an interactive experience for the customer, as well; it gives the customer a consistent view of the enterprise, which is critical in establishing relationships with your customers.

A Customer Event Hub in Action

For example, describes Prabhakar, a high-value banking customer is trying to transfer money online but cannot do it. As a result, the customer calls the bank’s technical support line. Unfortunately, this leads to even greater frustration.

Prabhakar suggests instead that financial institutions consider the possibilities of a call center response application that understands the needs of the caller. If the system knew, for example, what the customer wanted, it could route the caller to a much more immediate answer and result in a much more satisfying experience. “That’s the kind of use you can get from a Customer Event Hub,” he said.

Key Advantages for Your Business

According to Prabhakar, “All enterprises need to operate a CEH; it’s imperative for business agility as well as competitive advantage.” Some of the benefits of operating a CEH include:

Enhanced customer service and real-time personalization

“We all want the services and channels we engage with to be aware of who we are, what we like, and to respond accordingly,” said Prabhakar. “But there’s often a lag between when we exhibit certain behaviors and when the systems pick them up.” A CEH provides a way for enterprises to bridge that gap.

Innovative event-driven applications

As we’re increasingly finding new ways of engaging and working with the social channels, the CEH gives you the capability of building the next-generation infrastructure for new applications.

Security

Security is enhanced because the CEH lets you track up-to-the-minute activity on all your users and customers interacting with your enterprise.

Increased operational efficiency

With the CEH, you can eliminate the losses that are the result of a mismanaged application, mismanaged effort, or mismanaged expenses. This lowers the operational costs, which also means you increase the operational efficiency of the enterprise.

Now that we understand the purpose and benefits of CEHs, let’s take a look at how to build one.

Capturing and Integrating Events

According to Prabhakar, the first phase of enablement involves pulling together or capturing all the interaction channels and then integrating them. This means creating an event consolidation framework, often referred to as the event fire hose. This is how you bring the events into the CEH.

What kind of data and events are in the fire hose? Social media, structured and unstructured data, electronic files, binary files, teller notebooks, and so on—in other words, an ever-exploding and always expanding set of formats, both human- and machine-generated.

Naturally, due to the diversity of formats, you’re not going to have a uniform level of control over all of this data. “Your capability of running an application across all these channels will be limited by not being natively tied to those channels,” said Prabhakar. And this is what the CEH solves.

Sanitizing and Standardizing Events

Next, you need to sanitize and standardize the data. According to Prabhakar, “The goal is to create a consistent understanding of those events for your consuming endpoints.” An additional goal, of course, is to meet compliance and regulatory requirements. Ultimately though, standardization makes it possible for you to thread a story together across these channels and events.

Prabhakar explained that standardizing the data and preparing it for consumption primarily involves attaching metadata to every event. This process generally involves threading a handling mechanism around each event so that anybody can identify it, parse it out, and take action around it.

Delivering Data for Consumption

With the CEH, you can deliver data to various feeds and applications. According to Prabhakar, “If you’re delivering the data to an HBase cluster, chances are your online web applications could directly reference them and you can have it deliver these events in a very low latency manner.” Thus, you can access the data online across your enterprise. Prabhakar explained that you can also send this data into batch or offline processing stores.

In the earlier customer experience example, the call center application magically knew that a valuable customer had been trying to do something on the company’s website. It knows because the data has been delivered to another channel to produce a more meaningful user engagement.

Sounds relatively straightforward, doesn’t it? If so, why isn’t everyone building one?