Learning Library

← Back to Library

Securing Data While Running: Confidential Computing

Key Points

  • Confidential computing fills the missing “in‑use” security layer, protecting data while it’s being processed, complementing the existing at‑rest and in‑transit encryption paradigms.
  • The primary threats it addresses include malicious actors scraping data, memory‑dump attacks, insider threats, and the risk of exposing sensitive information to external partners or vendors during collaboration.
  • The technology relies on hardware‑based memory partitioning at the server level, creating isolated “enclaves” that keep data hidden even from the host operating system and privileged users.
  • IBM actively participates in the Confidential Computing Consortium, promoting cross‑industry collaboration to advance these hardware‑rooted solutions and integrate them into real‑world use cases.

Full Transcript

# Securing Data While Running: Confidential Computing **Source:** [https://www.youtube.com/watch?v=pMHxLBJ6_UA](https://www.youtube.com/watch?v=pMHxLBJ6_UA) **Duration:** 00:08:45 ## Summary - Confidential computing fills the missing “in‑use” security layer, protecting data while it’s being processed, complementing the existing at‑rest and in‑transit encryption paradigms. - The primary threats it addresses include malicious actors scraping data, memory‑dump attacks, insider threats, and the risk of exposing sensitive information to external partners or vendors during collaboration. - The technology relies on hardware‑based memory partitioning at the server level, creating isolated “enclaves” that keep data hidden even from the host operating system and privileged users. - IBM actively participates in the Confidential Computing Consortium, promoting cross‑industry collaboration to advance these hardware‑rooted solutions and integrate them into real‑world use cases. ## Sections - [00:00:00](https://www.youtube.com/watch?v=pMHxLBJ6_UA&t=0s) **Protecting Data While in Use** - Alex Greer introduces confidential computing as the missing third pillar—securing data during processing—to complement at‑rest and in‑transit encryption, highlighting cross‑industry collaboration and end‑to‑end security. - [00:03:20](https://www.youtube.com/watch?v=pMHxLBJ6_UA&t=200s) **Isolated Enclave Controls Access** - The passage describes an encrypted, physically isolated “enclave” that functions as a black box, granting decryption keys only to authorized programs while blocking insiders, malicious actors, and untrusted partners. - [00:06:36](https://www.youtube.com/watch?v=pMHxLBJ6_UA&t=396s) **Secure Multiparty Computing Use Cases** - The speaker outlines how multiparty computing enables safe data sharing across institutions, protects IP during collaborations, and mitigates insider threats by preventing exposure of sensitive information. ## Full Transcript
0:00How are you making sure that your highly sensitive  information is protected while you're running it? 0:06Hi my name is Alex Greer from the IBM Cloud  team, and make sure to like and subscribe. 0:10So, before I get started talking about  confidential computing I want to talk 0:13a little bit about why it's such an exciting  field. So, confidential computing among other 0:18reasons is exciting for the fact that we're seeing  cross-collaboration in the tech space to actually 0:24drive the technology forward, and so it's awesome  to see people reaching across a competitive 0:30aisle to do that, a lot of bright minds. The  second is that the technology is going to 0:34directly complement the existing data encryption  paradigm that we have today and make it an even 0:40more complete end-to-end story. So, before I get  into the actual technology and some of the use 0:45and value behind it let's start with the  existing pillars of data encryption today. 0:50So you start with protecting your data while  it's at-rest, so when you're storing it. 0:54So we've got at-rest, you can think of this as  whatever information you'd like to. We'll just 1:01represent it simply here. Now we have information  that we want to transit from point-to-point. So 1:08in order to do that securely we need to protect  it while it's in-transit, so we say in-transit. 1:19But what's missing in today's  story is this third pillar here. 1:23What are you doing to protect it  while you're actually running it? 1:28The groups you're going to have to protect  yourself against, one, are malicious actors 1:32who want to do things like scrape that data.  You're also going to want to protect against 1:37memory dumps, things of that nature. We have the  inevitable threat that we have to protect against 1:43which is insider threats. And, in addition,  we also have a lot of collaboration that we 1:49want to go on between us and either a trusted  vendor or even a trusted technology partner, 1:55but at the same time we really don't want to  expose a piece of highly sensitive information 1:59to them even though we want them to  be able to take advantage of it. So, 2:02we've got also our partners and vendors here. How  can we ensure that this information is not only, 2:12not visible to these parties, but also protected  from the worst case scenario? That's where 2:17confidential computing comes in. So, let's talk  about what confidential computing looks like. 2:25We earlier talked about the collaboration  across technology leaders in the space, 2:30we at IBM are a part of the Confidential Computing  Consortium. So the definition that we follow for 2:34confidential computing is as such, confidential  computing is a hardware-based technology that 2:39allows for the physical partitioning of memory  at the server level. So let's draw our stack. 2:48We've got our hardware level on the  bottom which is where the actual physical 2:52partitioning of the memory is going to take place, 2:55and then we have the middleware level, and then  for the example but not exclusive to we're going 2:59to talk about any containerized abstraction of  this. So we're going to have containers here. 3:11So what's taking place is that at the hardware  level we have that physical partitioning of the 3:16memory which allows for you to actually  run that application in its own silo. 3:20So the silo in the scenario that we've  painted here is going to be called an enclave. 3:26So we've got an enclave here, one, two,  three, etcetera, etcetera, etcetera, 3:30etcetera. So these enclaves can have applications  run in that that physically isolated environment, 3:37but let's take a look into more detail  about what that actual enclave is. 3:42So the enclave itself functions like a black  box so to speak. This black box or enclave 3:52has that data that we spoke about earlier we'll  make it the same one from our previous example, 3:58but what it also has in here are the set of  techniques or the things that or the actual 4:02processing and the procedures for that processing  that information so we've got our techniques here. 4:11What this system does in a scenario in  which we had the malicious actor from 4:16earlier, we had the insider threat,  and we even had our own partners -- 4:24what it does is it has an encryption key that  it only extends out to the authorized program. 4:32That allows for that authorized program to decrypt  the information running within this physically 4:37isolated silo and to be able to actually perform  its set of processes. So, this authorized program 4:44can do so, but that key is not extended to  the other parties. So that right privilege, 4:50since it's not offered here, prevents the access  from for a malicious actor. It prevents the access 4:57to an insider threat. And then finally,  it even prevents access to a partner. 5:02What actual access is it preventing now? This  is access to modify that code as well as and 5:07to view that data while it's actually inside this  physically partitioned silo. But what's really 5:14important for our design is that we verify  that the interaction with that code or data 5:20was what we hope for it to be. So  we need to have what we refer to as 5:24attestation reports. So attestation reports  you can see here, attestation reports. 5:33And just for good measure, we've got in here  that encryption key that we talked about. 5:40So let's get back to the key value proposition we  just discussed. So what this secure enclave this 5:46black box allowed us to do, it gave us a data  and code integrity that we didn't have before. 5:53One, it reduced the visibility of  that data while it was being run 5:58only to the the authorized program  itself. So, we have restricted visibility. 6:04The second, is that it took that data and  it actually prevented these parties from 6:09making any sort of undue modification  to the actual code itself. So no mod, 6:17or unauthorized mod, and then what we were able  to do was verify the actual interaction with that 6:23code and that data via attestation reports which  is very critical for corroborating the story 6:30that our system is telling us. So now let's  look at the use cases that have been enabled. 6:37So the first use case here we're going  to talk about is multiparty computing. 6:42Multiparty computing coming down to right here  we're actually working with the technology 6:47partner, it allows for us to take highly sensitive  information and exchange it with other parties 6:52without actually exposing it so therefore what we  can do is we can share data sets, we can actually 6:57collaborate, and perform functions commonly on top  of highly sensitive information. So you can think 7:04about maybe a collaboration between two research  institutions who otherwise in the past had to 7:08go around a very complicated path for actually  getting approval to exchange that information, 7:13now they've got that quick and easy path  by simply not allowing that information to 7:18be exposed to the other party. The next case  that we're going to talk about is IP. So IP, 7:23you can think of that next great discovery  that your company is making perhaps you've 7:28discovered some sort of pharmaceutical  innovation and you have the actual, 7:33the blend or the chemical composition and the  other things that are unique about your solution 7:37and you're wanting to actually share that with  another party, but again without exposing that 7:41information to them. So this is another  great scenario where you're protected. 7:46Now the final one, it's a real situation that we  all have to tackle today is that we give the keys 7:52to the kingdom essentially to the people that we  trust and we hire, but an insider thread is always 7:57a possibility. So we have to protect our workloads  from that possibility. So we have insider threats 8:03here and now we've protected ourselves from even  the case in which our own turn on us. So now 8:09that we've got the use cases laid out you can see  the clear value that is provided by confidential 8:13computing. Confidential computing is focused on  protecting application data while you're actually 8:19running it, and it allows for us to be able to  collaborate more freely with other parties as well 8:26as protect ourselves in a new way from malicious  actors whether external or internal. Thank you for 8:31listening. If you have questions please drop us  a line below. If you want to see more videos like 8:35this in the future, please like and subscribe.  And don't forget, you can grow your skills and 8:39earn a badge with IBM CloudLabs which are free  browser-based interactive Kubernetes labs.