Page 97 - Cyber Defense eMagazine September 2023

P. 97

Meeting that challenge has been the driving force for Confidential Computing. At a high level, Confidential
Computing protects data in use during processing, with sensitive data isolated in the CPU and encrypted
in memory while it’s processed. The main premise behind separating user data from platform owner
control (whether through a service or on-prem) is that those controlling the platform and those accessing
data on the platform being processed are two separate entities.

Confidential Computing delivers a variety of security benefits associated with hardware-based
enforcement of data and IP access policies. For example, protecting against un-permissioned data
access (even with escalated privileges), containing the “blast radius” of compromised software or
unauthorized users, and protecting data and IP deployed at partner or remote sites. But Confidential
Computing is also transforming businesses by helping to deliver a platform for controlled, compliant
activation of sensitive data. That might include compliant multi-party or inter-department collaboration,
privacy-preserving data services, confidential and compliant AI, and more.

In order for this technology to become more pervasive, it must be usable, deployable, and not detrimental
to either the data owner’s or the platform owner’s cost models – all while delivering security value. Finding
this balance is what both drives and constrains progress in this field. What are some of the new security
challenges the industry faces as we continue to advance in Confidential Computing? Here are four:

• Post-Quantum Crypto (PQC) Hardening – The entire computing industry is amid what will be a
decade-long transition (or more) to post-quantum safe computing. This transition is where the
cryptographic algorithms used in the modern era must change to prevent attacks from quantum
computer-based attacks. In some instances, the transition efforts may require that we encrypt
data with a change of key length. However, in other cases, new PQC algorithms need to be
invented and standardized. There is a lot of working happening in this area.

• Attestation Complexities – Part of the element of a Confidential Computing model is
understanding the state of the security posture of platforms to which execution is deployed. This
process is known as attestation, and it is still in its infancy. Developing and appraising policies to
determine actions when platforms are out of update or configuration isn’t up to standard is still a
developing area. Finding ways of making these processes more understandable, digestible, and
automated continue to develop.

In addition, we have seen many users of Confidential Computing request heterogeneity in the
platforms, clouds, and software stacks. This introduces other challenges in making cross-
platform, cross-cloud, and cross-trust authority assessments more complex.

• Side-Channel Attacks – A side-channel attack, or an attempt to extract secrets from a chip or a
system, can happen in any digital system. For example, CPU vulnerabilities (such as Meltdown
and Spectre) allowed unauthorized reads of data in running programs and illustrated the
challenges of side-channel attacks. While many side-channel attacks are shown in academic or
laboratory scenarios, the risks are increased in cloud computing environments that rely on co-

92 93 94 95 96 97 98 99 100 101 102