Transition to VaaS part 4

cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Announcements

Welcome Idira to LIVEcommunity!

Our community just got stronger! Explore new identity security resources, connect with experts, and discover more ways to learn and collaborate.

Meet Idira

Transition to VaaS part 4

L1 Bithead

 

 

I recently started deploying vsatellites around our environment, and I am in the process of finishing 12 deployments, across two cloud providers in 3 different regions each. The logic behind this kind of deployment is because of the DEK the vsats store. Venafi highly recommends, and I agree with this, to have more than one single vsat out there. If you only have one vsat and it goes down, you lose your DEK, unless you have backed up the DEK somewhere. This makes sense.

 

But new and interesting problems come up when you are in a hybrid environment or all cloud environment: putting all of your vsats in one cloud provider and one region can cause its own problems - total system failure if that region goes down for any reason. And this has happened at least twice in the case of AWS in the last couple of years.

 

So, we are deploying a dozen vsats across half a dozen regions in two different cloud providers. Nice, safe, redundant.

 

The other reason why we are installing so many is because we want to scan our edge for previously unknown certificates or installed locations. This came up recently when I received an alert from our IR team about a certificate that was about to expire in one day on TCP25, and that app team (and Venafi) said the cert had been rotated to the newest version.

 

Now, Venafi will tell you that this is a bad idea. For varying reasons, it can be a bad idea, and to do it efficiently requires a significant level of planning. Venafi believes that everyone has a solid grasp on their inventory, this is the first two steps of the top 18 critical security controls (formerly the top 20), but anyone who has been in IT or Security for any length of time knows that, frankly, app and dev teams never state absolutely everything that they are doing.

 

Vsats are limited in capability. Vsats and Scanafi (if you're using this method) are limited to two million scannable sockets. This seems in line with the other rather absurd design decisions around these tools, but it is what it is. Regardless, this scanning limit means that math needs doing. If you're a large company, this could be a significant endeavor. But some fast calculator skills can get me into the ballpark. To do this, I copied my discovery targets out to an excel spreadsheet, deleted extraneous data, and counted my rows. Many of these are subnets rather than individual hosts, that I had to keep in mind doing this. Opened calculator and started mathing. Pretty quickly I was able to break down my number of rows into an even set that would allow to reach about 1.8M sockets per engine. I felt like this would allow room for the subnets easily without going over the 2M threshold. This is also important because Venafi states that this should be able to complete in 24 hours.

 

Installation of vsats is limited to 2 operating systems: RHEL and Ubuntu. When I asked about Centos I got hemmed and hawed about it. Not real clear why they only want to support these two variants, but again, it is what it is.

 

Officially, my company does not support either of these operating systems. When I asked for an Ubuntu box, I was told no. It seems strange that these would be limited to officially supporting only these two types of operating systems, and they are not even the same core versions of Linux, but hey - fun times.

 

So, I requested my own subscriptions, and built them myself. To keep things easy, I used the CIS Level 1 Ubuntu 20.04 image in the marketplaces. In Azure, this works fine. Install the vsats, configured the scans, and all was well with the world.

 

In GCP, not so much. Officially, Venafi does not support GCP. But it does appear to work, if you want to fight a small war with the vsat installation.

 

It takes three attempts to install vsatellite on an Ubuntu hardened image in GCP.

 

The first attempt, it fails at the bootstrapping step, but it DOES check in with Venafi and adds it to the vsat screen. Delete from venafi, and run ./vsatctl uninstall - and it fails. Somewhere in the process, vsat puts a vsatctl.lock file in the /tmp directory. Delete this file and run ./vsatctl uninstall --force and it will remove it. You have to use the --force option or it will fail to uninstall.

 

The second attempt fails at about 90% completion of vsatellite. No idea why it gets this far and fails, but it does. Rinse and repeat the above process: delete the vsatctl.lock file, uninstall with force, and try again.

 

The third time it works perfectly. Finishes the installation, appears in the vsat screen, and appears in the discovery menu and scans can be configured.

 

Now, why Venafi does not officially support GCP, one of the three biggest cloud providers, I don’t know. I feel like this is a decision by some business guy who thinks that this is a smart idea. It wasnt.

 

The other problem with vsatellites: they CANNOT be installed in a kubernetes service. They HAVE to be installed on an operating system, which then installs Kubernetes and the vsat service. Why? I have no idea. But it is not a great concept considering the proliferation of containers, and that the big 3 cloud providers all support some form of containerization as a service.

 

Being able to install these in a container service, whether Azure Kubernetes Service (AKS), or Googles Kubernetes Engine (GKE) would be the MUCH smart and easier way to do this. It would save on resources and cost to us and allow for on the spot expansion and contraction of system resources. Again, I feel like this was some suit calling the shots with little understanding of what they were doing.

 

In the end, the scanning engines are running, and I have found dozens of previously unknown certs, self-signed certs on the edge, and several certificates on a LOT of ports other than 443.

 

The lesson here is programmers are not always 100% honest with security teams about what they are doing and scanning the edge for the unknown can give you great insight into what might get missed in Machine Identity Management

 

Now if I can just get Venafi to grab the TTLs out of the TCP handshakes and post probable OS in certificate inventories next to IPs, I might be on Cloud 9. Since this is data that they already have, it shouldn't be a huge deal to get this data in there.

-----

[Moderator Note: Note: To maintain clarity for the community, the links originally included in this post have been removed as they are no longer active.]

0 REPLIES 0
  • 50 Views
  • 0 replies
  • 0 Likes
Like what you see?

Show your appreciation!

Click Like if a post is helpful to you or if you just want to show your support.

Click Accept as Solution to acknowledge that the answer to your question has been provided.

The button appears next to the replies on topics you’ve started. The member who gave the solution and all future visitors to this topic will appreciate it!

These simple actions take just seconds of your time, but go a long way in showing appreciation for community members and the LIVEcommunity as a whole!

The LIVEcommunity thanks you for your participation!