Skip to content
View poojakira's full-sized avatar
💭
open to work
💭
open to work

Highlights

  • Pro

Block or report poojakira

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
poojakira/README.md

Pooja Kiran - ML Security Engineer

I build security controls around AI/ML systems: model supply-chain checks, LLM/agent security monitoring, adversarial robustness experiments, dataset-poisoning detection, model privacy evaluation, and secure ML serving.

LinkedIn Email


Selected Public Work

Project What it demonstrates Status
hf-model-provenance-scanner Hugging Face model provenance, impersonation, pickle-risk, and supply-chain signal checks. Public repo
mcp-security-gateway-monitor MCP tool-call monitoring for prompt injection, PII leakage, shadow servers, and exfiltration patterns. Public repo
llm-redteam-framework Adversarial prompt generation and offline detector experiments for LLM red-team workflows. Public repo
dataset-poisoning-detector Dataset poisoning and anomalous-sample detection experiments. Public repo
model-privacy-attacks Membership-inference and model-privacy attack evaluation. Public repo
adversarial-ml-lab FGSM, PGD, and C&W adversarial ML attacks plus adversarial-training defenses. Prototype
PulseNet-RUL-Forecasting RUL forecasting and FDIA attack verification. Public repo

Technical Focus

  • Model supply chain: provenance checks, pickle-risk detection, typosquat/homoglyph signals, SBOM-oriented workflows
  • LLM and agent security: prompt-injection monitoring, tool-call inspection, exfiltration and PII-leak detection
  • Adversarial ML: FGSM, PGD, C&W attacks; adversarial training; CI-oriented robustness checks
  • Data security: dataset-poisoning detection and per-sample anomaly attribution
  • Secure serving: JWT auth, RBAC, rate limiting, audit logging, Prometheus metrics
  • Stack: Python 3.11+, PyTorch, scikit-learn, FastAPI, Docker, GitHub Actions

Research and Publications

Cybersecurity Innovation Researcher - TEM 598 Technology Innovation Lab, Arizona State University x Honeywell Aerospace Innovation Hub. Contributed to a graduate research practicum exploring AI and cybersecurity challenges for aerospace systems.


Availability and Contact


Last updated: July 2026. Claims are limited to public, inspectable work.

Operational Runbook

Verified profile-maintenance path uses RUNBOOK.md. Minimal validation:

py -3.12 -m pip install -r requirements-dev.txt
py -3.12 -m pytest tests -q

Scope note: this profile README links to public work; claims should remain limited to inspectable repositories and verified artifacts.

Pinned Loading

  1. hf-model-provenance-scanner hf-model-provenance-scanner Public

    Scan Hugging Face model repos for provenance, impersonation, pickle-risk, and supply-chain signals.

    Python

  2. mcp-security-gateway-monitor mcp-security-gateway-monitor Public

    Monitor MCP tool calls for prompt injection, PII leakage, shadow servers, and exfiltration patterns.

    Python

  3. llm-redteam-framework llm-redteam-framework Public

    Generate adversarial prompts and evaluate an offline detector for LLM red-team experiments.

    Python