metadata
base_model: google/gemma-4-e2b-it
tags:
- text-generation-inference
- transformers
- gemma4
- peft
- lora
- cybersecurity
- macos
- privilege-escalation
- tcc-bypass
- sip
- apple-security
license: apache-2.0
language:
- en
Gemma 4 E2B — macOS Privilege Escalation Expert
A QLoRA fine-tuned version of Gemma 4 E2B Instruct specialized in macos privilege escalation. Specialized in macOS privilege escalation: SIP bypass, TCC bypass, LaunchDaemon misconfigurations, dylib injection/hijacking, Keychain attacks, and macOS security hardening.
Part of the rezaduty cybersecurity model family.
Expertise
- macOS security model: SIP, TCC, Gatekeeper, XPC, Sandbox
- SIP bypass: boot arguments, rootless.conf, third-party kext loading
- TCC bypass: SQLite injection, electron app abuse, XPC service exploitation
- LaunchDaemon/LaunchAgent misconfigurations and PLIST injection
- dylib hijacking and injection via DYLD_INSERT_LIBRARIES
- Keychain extraction: security CLI, chain-break attacks
- macOS CVEs: CVE-2022-22583 (powerdir), CVE-2021-30892 (shrootkit)
Model Details
| Property | Value |
|---|---|
| Base model | google/gemma-4-e2b-it (2B parameters) |
| Fine-tuning method | QLoRA (rank 16, α 16) |
| Domain | macOS Privilege Escalation |
| Dataset | rezaduty/cybersecurity-qa-v2 |
| License | Apache 2.0 |
Usage
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch
base_model = "google/gemma-4-e2b-it"
adapter = "rezaduty/gemma4-e2b-privesc-macos"
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 macOS privilege escalation and security. Provide deep technical answers on macOS privesc techniques, TCC bypass, SIP, macOS security internals, and hardening with specific commands, tool names, and CVE references."}]},
{"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 macOS privilege escalation and security. Provide deep technical answers on macOS privesc techniques, TCC bypass, SIP, macOS security internals, and hardening with specific commands, tool names, and CVE references.