WHISPER
A Generalizable Framework for Securely Querying Public Data
WHISPER is a team research project building a generalizable framework for securely querying public data sources. This case study describes my individual contribution — the ephemeral compute and identity layer — not the full system or other contributors' work.
Project overview
WHISPER's goal, as a team, is to let services query public data sources without embedding long-lived credentials in client-facing code or maintaining always-on infrastructure that expands the attack surface over time.
My role on the team was the ephemeral compute and identity layer underneath that goal: provisioning short-lived, scoped AWS compute on demand, and making sure it never held more access — or existed longer — than a given job required.
Ephemeral compute & identity layer
Six pieces of infrastructure I built and integrated for the query-execution path.
Ephemeral compute provisioning
Used EC2 Launch Templates to spin up short-lived compute instances scoped to a single query job, rather than relying on persistent, always-on servers.
Automated lifecycle management
Boto3-driven automation to launch, monitor, and tear down compute resources automatically once a job completed — no manual cleanup, no orphaned instances.
Role-based authentication
IAM roles scoped per job rather than static, shared credentials, so each compute instance only had the permissions its specific task required.
Removal of embedded credentials
Replaced hard-coded and embedded credentials in client-facing code with temporary, role-assumed credentials issued at runtime.
API request layer
A FastAPI service behind AWS API Gateway and Lambda, handling request routing and triggering ephemeral compute provisioning for incoming query jobs.
Secure distributed workloads
Coordinated compute across distributed, ephemeral instances while keeping access control consistent across the fleet.
Request-to-teardown flow
How a query job moves through the infrastructure I built, from incoming request to automatic cleanup.
Step 1Request received
Step 2Job authorized
Step 3Ephemeral compute provisioned
Step 4Query executes
Step 5Automatic teardown
Why this approach
Long-lived servers and long-lived credentials both accumulate risk over time — every day they exist is another day they can be misconfigured, leaked, or exploited. Building the compute layer to be ephemeral by default and identity to be scoped and temporary shrinks that window deliberately: infrastructure that exists only for the duration of a job, with access limited to exactly what that job needs.
Publication
This work was presented as a research abstract and poster at the HBCU/MI Summer Research Program Symposium, Air Force Research Laboratory, August 2026, as a co-author and presenter. Full citation details are on the Publications page.