guest@manav-thakkar:~/operations
OPERATIONS _
Accessing project database... Displaying classified operations and research initiatives.
Many Looks Beat One Look: Commercial LLM Guardrails vs. Multi-Turn Attacks
Research Paper · ACDSA 2027 (Submitted)
$ cat project_summary.txt
Measured four commercial guardrails (AWS Bedrock, Azure Prompt Shields, Azure Content Safety, Google Vertex Model Armor) on 2,489 multi-turn attacks. Showing the guardrail the whole conversation never improves conversation-level detection, and 13% of attacks pass all four products.
AI Red Teaming
LLM Guardrails
Multi-Turn Jailbreaks
Python
Security Measurement
GenAI Compliance Mapping & Action Planner
$ cat project_summary.txt
Full-stack app that maps controls across ISO/IEC 27001, NIST 800-53, CIS Controls, SOC 2, PCI DSS and GDPR using embedding-based similarity search, then generates actionable compliance to-do lists with GPT-4o. Includes sorting, filtering and CSV export.
FastAPI
React
TypeScript
OpenAI Embeddings
GRC
WiFi Deauth and Cloning using ESP8266
IEEE-ICAST 2022 · ID: CS_122
$ cat project_summary.txt
Research project on WiFi security vulnerabilities, presented at IEEE-ICAST 2022. The project demonstrates deauthentication attacks and cloning techniques using inexpensive ESP8266 hardware.
ESP8266
WiFi Security
Penetration Testing
IEEE 802.11
Secure AI Chatbot
$ cat project_summary.txt
An Ethical Input Checker: a web application designed to screen interactions with Large Language Models (LLMs) for potentially problematic content.
JavaScript
LLM Security
Red Teaming
OWASP LLM Top 10
FLAG CAPTURED: Project intelligence successfully acquired.
Continue exploring to gather more intelligence...