In the introduction to our series on Monitoring the Hybrid Cloud we went through all the disruptive forces which are increasingly complicating security monitoring. These include the accelerating move to cloud computing and expanding access via mobile devices. Those new models require much greater automation, and significantly less visibility and control over the physical layer of the technology stack. So you need to think about monitoring a bit differently.

This starts with getting a handle on the nuances of monitoring, depending on where applications run, so we will discuss monitoring both IaaS (Infrastructure as a Service) and SaaS (Software as a Service). Not that we discriminate against PaaS (Platform as a Service), but it is similar enough to IaaS that the concepts presented are similar. We will also talk about private clouds because odds are you haven’t been able to unplug your data center, so you need to provide an end-to-end view of the infrastructure you use, including both technology you control (in your data center) and stuff you don’t (in the cloud).

Monitoring IaaS

The biggest and most obvious challenge in monitoring Infrastructure as a Service is the difference in visibility because you don’t control the physical stack. So you are largely restricted to logs provided by your cloud service provider. We see pretty good progress in the depth and granularity available from these cloud log feeds, but you still get much less detail than from devices in your data center.

You also cannot tap the network to get actual traffic (packet capture). IaaS vendors offer abstracted networking so many networking features you have come to rely on aren’t available. Depending on the maturity of your security program and incident response process, you may not be doing much packet capture on your environment now, but either way it is no longer an option now in the cloud.

We will go into more detail later in this series, but one workaround is to run all traffic through a cloud-based choke point for collection. In essence you perform a ‘man-in-the-middle’ attack on your own network traffic to regain a semblance of the visibility inside your own data center, but that sacrifices much of the architectural flexibility drawing you to the cloud in the first place.

You also need to figure out where to both aggregate collected logs (both from the cloud service and from specific instances) and where to analyze them. These decisions hinge on a number of factors including where the technology stacks run, the kinds of analysis to perform, and what expertise is available on staff. We will tackle specific architectural options in our next post.

Monitoring SaaS

If monitoring IaaS offers a ‘foggy’ view compared to what you see in your own data center, Software as a Service is ‘dark’. You see what the SaaS provider shows you, and that’s it. You have access to neither the infrastructure running your application, nor the data stores that house your data. So what can you do?

You can take solace in the fact that many larger SaaS vendors are starting to get the message from angry enterprise clients, and providing an activity feed you can pull into your security monitoring environment. It won’t provide visibility into the technology stack, but you will be able to track what your employees are doing within the service – including administrative changes, record modifications, and login history.

Keep in mind that you will need to figure out thresholds and actions to alert on, mostly likely by taking a baseline of activity and then looking for anomalies. There are no out-of-the-box rules to monitor SaaS. And as with IaaS you need to figure out the best place to aggregate and analyze data.

Monitoring a Private Cloud

Private clouds virtualize your existing infrastructure in your own data center, so you get full visibility, right? Not exactly. You will be able to tap the network within the data center for additional visibility. But without proper access and instrumentation within your private cloud you cannot see what is happening within the virtualized environment.

As with IaaS, you can route network traffic within your private cloud through an inspection point, but again that would reduce flexibility substantially. The good news is that many existing security monitoring platforms are rapidly adding the ability to monitor within virtual collection points which run in a variety of private cloud environments. We will address alternatives to extend your existing monitoring environment later in this series.

SLAs are your friend

As we teach in the CCSK (Certificate of Cloud Security Knowledge) course, you really don’t have much leverage to demand access to logs, events, or other telemetry in a cloud environment. So you will want to exercise whatever leverage you have during the procurement process; document specific logs, access, etc. in your agreements. You will find that some cloud providers (the smaller ones) are much more willing to be contractually flexible than the cloud gorillas. So you will need to decide whether the standard level of logging from the big guys is sufficient for the analysis you need.

The key is that once you sign an agreement, what you get is what you get. You will be able to weigh in on product roadmaps and make feature requests, but you know how that goes.


If a large fraction of your technology assets have moved into the cloud there is a final use case to consider: moving the collection, analysis, and presentation functions of your monitoring environment into the cloud as well. It may not make much sense to aggregate data from cloud-based resources, and then move the data to your on-premise environment for analysis. More to the point, it is cheaper and faster to keep logs and event data in low-cost cloud storage for future audits and forensic analysis.

So you need to weigh the cost and latency of moving data to your in-house monitoring system against running monitoring and analytics in the cloud, in light of the varying pricing models for cloud-based versus on-premise monitoring.

But the reality is that you are likely to run a hybrid for a while, with infrastructure in both places (cloud and on-premise data center) for the foreseeable future. As we mentioned above, there are a bunch of decision points to figure out whether SOC systems should run in the cloud, on-premise, or both. Our next post will dig into possible architectures for cloud-based collection, and a variety of hybrid SOC models.