How i jailbreaked a 14B parameter Ai Model
Summary A single-layer regex input filter on a locally-hosted Qwen2.5-14B-Instruct chatbot was bypassed using a fake system-message injection combined with explicit refusal-suppression instructions, p
My Space
just a tech!!!
Latest articles
Quantifying Prompt Injection Resistance in Small Language Models (7B parameter model)
Abstract / TL;DR Over two weeks, I deployed and red-teamed two locally-hosted Qwen2.5 models (0.5B and 7B parameters) behind a custom three-layer guardrail stack (regex input filtering, system-prompt
Quantifying Prompt Injection Resistance in Small Language Models
Methodology Target systems: Qwen2.5-0.5B-Instruct and Qwen2.5-7B-Instruct, run via Ollama (local, CPU) and Hugging Face transformers with 4-bit quantization (CUDA, Colab T4), respectively. Guardrail a
Getting a Local Model Running
Before any Python script can talk to a model, the model itself has to be running and listening for requests. This is the part that happens entirely outside your code — in the terminal, one time (or on
Running Any Local Model Without the OpenAI Library
using the OpenAI SDK against Ollama works, but it hides how the actual request/response cycle works, and it implies you need OpenAI's library at all, which you don't. Here's the same chatbot logic usi
Building Your Own Guarded Chatbot via API : A Line-by-Line Walkthrough
If you already have a model pulled locally (via Ollama, LM Studio, or similar), you're one script away from a working chatbot. But a guarded chatbot — one that resists jailbreaks and stays on-topic —
LLM Token ID Vs Embedding
1. Core Definitions Token ID: A unique, arbitrary integer — essentially a barcode (e.g., "cat" = 1042). It carries zero conceptual meaning on its own; it's just a lookup key, used to keep the system m
IAM Fundamentals: OAuth, JWT, Least Privilege, and MFA — How Identity Trust Actually Works
1. OAuth — Delegated, Scoped Authorization The problem it solves: Before OAuth, if an app like "PhotoPrint" needed access to your Google Photos, you'd hand over your actual Google password. That's a m
Breaking JWT Trust: The RS256 → HS256 Algorithm Confusion Attack
The Setup JWTs (JSON Web Tokens) are self-contained, signed claims. A server issues a token like: { "header": { "alg": "HS256", "typ": "JWT" }, "payload": { "user": "alice", "role": "user", "exp":
