I've done a ton of job before New Year, on Enclaves Framework, and CDK Dev Stack (former CDK SOA Backend), closed most of the tech debt.
Made new init system on Rust (systemd inspired) for enclaves internal provisioning (services, processes).
Started Enclave's Engine development. This component is for Enclave's provisioning on host. (Think of it like Docker Engine with API, Docker Compose, YAML configurations, and containerd runtime, but for secure enclaves.) First iteration already published.
For now, Enclaves Framework is a turn key solution for AWS Nitro Enclaves, for making custom Nitro Enclave images (with custom kernel, init, SLC, proxies, attestation server, and other components) with reproducible builds (supply chain security).
With Enclaves Engine there's a goal to make the same level of usability for confidential VMs, based on KVM, QEMU and Firecracker VMM (think of it as of your own self-hosted Enclaves platform as turn-key solution).
So, delivering Docker like developer experience for Enclaves - this motto is evolving by recent efforts! 🙌
https://github.com/sentient-agi/Sentient-Enclaves-Framework
Some of my experiments will be here in my own profile:
https://github.com/andrcmdr/secure-enclaves-framework
https://github.com/andrcmdr/cdk-dev-stack
Covering everything with exhaustive comprehensive documentation - documentation amount (in lines) exceeded the amount of code already! That's funny! 😁
Refactored main components - Pipeline Secure Local Channel protocol (through VSock) client-server implementation, VSock TCP set of proxies, and Remote Attestation Web Server - made proper error handling and structural logging with tracing for all components, made dynamic VSock buffers allocation for Pipeline SLC, refactored the RA Web-Server to make it modular.
Published paper about multi-hop re-encryption and delegated decryption, about cryptography difficulties for content protection and DRM in application to AI content producers and consumers (for community driven AI).
And published another paper about GPU TEE, attestation, coherent and unified memory, and how it's cause current scalability difficulties for TEE systems.
https://github.com/sentient-agi/Sentient-Enclaves-Framework/blob/main/docs/multi_hop_reencryption.md
https://github.com/sentient-agi/Sentient-Enclaves-Framework/blob/main/docs/multi_hop_reencryption_for_data_protection.proto.rs.md
https://github.com/sentient-agi/Sentient-Enclaves-Framework/blob/main/docs/unified_vs_discrete_memory_for_confidential_ai_and_cvms.md
https://github.com/sentient-agi/Sentient-Enclaves-Framework/blob/main/docs/unified_vs_discrete_memory_for_confidential_ai_and_cvms_2nd_iteration.md
If some of these sparkling your interest - give me a hint and text me! I'm looking for the TEE companies, who will also adopt and use Enclaves Framework and Enclaves Engine.
I think to provide a container like (Docker grade) developer and user experience for enclaves (hardware isolation and memory encryption) technologies for AI and crypto apps and lowering the entry barrier to hardware isolation technologies - is a great mission and ultimate data security goal (especially in context of cryptography and secrets in-memory protection) for the upcoming decade.
So, feel free to reach me if this is interesting for you as well!
#Enclaves
#TEE
#AI
#Cryptography
#Crypto
No reviews yet.