Preventing Model Inversion and Membership Inference in Deployed APIs
Defending APIs requires layered controls at training, serving, and boundary levels working together.
Anouk Mercer
Editor at Large
Anouk Mercer
Anouk Mercer covers zero-trust and provable data privacy, confidential computing on gpus and attestation and verifiable computing for Private AI Review.
13 stories
Cryptographic Proof of Data Deletion for AI Training Pipelines
Anouk Mercer
Confidential Federated Learning With GPU TEEs
GPU TEEs push confidentiality into the compute layer where model training actually happens.
Anouk Mercer
Confidential Computing Attestation vs TLS Certificate Trust Models
Attestation proves what code is running; TLS only proves who you're talking to.
Anouk Mercer
Configuring Local LLM Inference with Ollama for Data-Sensitive Workflows
Learn the configuration settings that keep sensitive data on your own machine.
Anouk Mercer
Preventing Sensitive Data Leakage in Enterprise AI Copilot Configurations
Audit permissions and classify data before deploying AI copilots to your organization.
Anouk Mercer
Homomorphic Encryption Schemes for ML Inference Comparison
Choosing the right homomorphic encryption scheme determines your ML inference speed and cost.
Anouk Mercer
AWS Nitro Enclaves for Sensitive Data Processing
Protecting sensitive data by isolating cryptographic operations from the host operating system.
Anouk Mercer
Intel SGX Enclave Development with the Open Enclave SDK
How to design SGX enclaves that don't leak secrets at the boundary.
Anouk Mercer
TEE TCB Updates and Patch Management in Production
SVN mismatches between patches and attestation can silently degrade production deployments.
Anouk Mercer
Sealing and Secret Provisioning to SGX and TDX Workloads
How SGX encrypts secrets for storage and migration across hardware boundaries.
Anouk Mercer
Intel TDX vs AMD SEV-SNP Architecture Differences
AMD distributes security across hardware layers; Intel centralizes it in firmware.
Anouk Mercer
Zero-Trust Architecture for AI Inference Infrastructure
Workload identity and continuous verification replace network perimeters in distributed AI systems.
Anouk Mercer