README.md

---
base_model: google/gemma-4-e2b-it
tags:
- text-generation-inference
- transformers
- gemma4
- peft
- lora
- cybersecurity
- active-directory
- red-team
- pentesting
- cybersecurity
- windows-security
license: apache-2.0
language:
- en
---

# Gemma 4 E2B — Active Directory & Red Team Expert

A QLoRA fine-tuned version of [Gemma 4 E2B Instruct](https://huggingface.co/google/gemma-4-e2b-it) specialized in **active directory & red team**.
Specialized in **Active Directory and red team techniques**: Kerberoasting, Pass-the-Hash, DCSync, BloodHound analysis, GPO abuse, and defensive hardening strategies.

Part of the [rezaduty cybersecurity model family](https://huggingface.co/rezaduty).

---

## Expertise

- Kerberoasting, AS-REP Roasting, and Kerberos delegation abuses
- Pass-the-Hash, Pass-the-Ticket, and credential harvesting
- DCSync, Golden/Silver Ticket attacks and detection
- BloodHound/SharpHound attack path analysis
- GPO abuse, ACL misconfigurations, and AdminSDHolder
- AD hardening: tiering model, PAW, LAPS, Protected Users group

---

## Model Details

| Property | Value |
|---|---|
| **Base model** | google/gemma-4-e2b-it (2B parameters) |
| **Fine-tuning method** | QLoRA (rank 16, α 16) |
| **Domain** | Active Directory & Red Team |
| **License** | Apache 2.0 |

---

## Usage

```python
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

base_model = "google/gemma-4-e2b-it"
adapter    = "rezaduty/gemma4-e2b-redteam-activedirectory"

tokenizer = AutoTokenizer.from_pretrained(adapter)
model = AutoModelForCausalLM.from_pretrained(
    base_model, torch_dtype=torch.bfloat16, device_map="auto"
)
model = PeftModel.from_pretrained(model, adapter)

messages = [
    {"role": "system", "content": [{"type": "text", "text": "You are an expert in Active Directory security and red team operations. You provide deep answers on AD attack paths, lateral movement, credential theft, and defensive hardening."}]},
    {"role": "user",   "content": [{"type": "text", "text": "Your question here"}]},
]
inputs = tokenizer.apply_chat_template(
    messages, tokenize=True, add_generation_prompt=True, return_tensors="pt"
).to(model.device)
output = model.generate(inputs, max_new_tokens=512, temperature=0.7, top_p=0.9)
print(tokenizer.decode(output[0][inputs.shape[-1]:], skip_special_tokens=True))
```

---

## System Prompt

```
You are an expert in Active Directory security and red team operations. You provide deep answers on AD attack paths, lateral movement, credential theft, and defensive hardening.
```

---

## See Also

- [General cybersecurity model](https://huggingface.co/rezaduty/gemma4-e2b-cybersecurity-interview) — full 646-example dataset
- [Docker & Container Security](https://huggingface.co/rezaduty/gemma4-e2b-docker-container-security)
- [Kubernetes Security](https://huggingface.co/rezaduty/gemma4-e2b-kubernetes-security)
- [AI & LLM Security](https://huggingface.co/rezaduty/gemma4-e2b-ai-llm-security)
- [All rezaduty models](https://huggingface.co/rezaduty)