rezaduty/gemma4-e2b-cybersecurity-interview

Gemma 4 E2B — Cybersecurity Interview Expert

A QLoRA fine-tuned version of Gemma 4 E2B Instruct specialized in deep, production-level cybersecurity knowledge. This model answers technical security interview questions with precision, concrete examples, and actionable recommendations.

---

Model Details

Property	Value
Base model	google/gemma-4-e2b-it (2B parameters)
Fine-tuning method	QLoRA (rank 16, α 16)
Trainable parameters	31M / 5.15B (0.60%)
Training data	646 curated cybersecurity interview Q&A pairs
Epochs	3
Final training loss	0.574
License	Apache 2.0

---

Expertise & Capabilities

This model demonstrates expert-level knowledge across the full spectrum of modern cybersecurity:

Cloud & Container Security

• Docker security hardening (rootless containers, capabilities, seccomp, AppArmor)
• Kubernetes RBAC, Pod Security Standards, network policies, admission controllers
• AWS IAM least-privilege design, ECR image scanning, Terraform security patterns
• Cloud-native threat modeling and attack surface reduction

DevSecOps & CI/CD

• Secure pipeline design (ArgoCD, GitHub Actions, GitLab CI)
• Supply chain security: SLSA, SBOM, sigstore/cosign, dependency verification
• Secrets management (Vault, AWS Secrets Manager, SOPS)
• Infrastructure-as-Code security scanning (Checkov, tfsec, Terrascan)

Application & Secure Coding

• OWASP Top 10 — root cause analysis and remediation
• Injection attacks (SQL, command, LDAP, template), XSS, SSRF, deserialization
• Authentication & authorization: OAuth 2.0, OIDC, JWT, PKCE, session security
• Cryptography: TLS configuration, key management, algorithm selection

Threat Intelligence & Offensive Security

• SOC operations, SIEM correlation rules, threat hunting
• MITRE ATT&CK mapping and adversary emulation
• Active Directory attack paths (Kerberoasting, Pass-the-Hash, DCSync)
• Red team tactics and purple team collaboration

Emerging & Specialized Domains

• AI/LLM security: prompt injection, model poisoning, guardrail bypasses
• OT/ICS/SCADA security: Purdue model, IEC 62443, air-gap strategies
• Blockchain & smart contract auditing (reentrancy, overflow, access control)
• Digital forensics, incident response, and malware analysis

---

Usage

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

base_model = "google/gemma-4-e2b-it"
adapter    = "rezaduty/gemma4-e2b-cybersecurity-interview"

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 cybersecurity engineer specializing in DevSecOps, "
            "container security, and cloud-native security. Answer technical interview "
            "questions with depth, precision, and concrete examples."
        )}]
    },
    {
        "role": "user",
        "content": [{"type": "text", "text": "Explain why running Docker containers as root is a security risk and how to fix it."}]
    },
]

inputs = tokenizer.apply_chat_template(
    messages,
    tokenize=True,
    add_generation_prompt=True,
    return_tensors="pt",
).to(model.device)

output = model.generate(
    input_ids=inputs,
    max_new_tokens=512,
    temperature=0.7,
    top_p=0.9,
    use_cache=True,
)
print(tokenizer.decode(output[0][inputs.shape[-1]:], skip_special_tokens=True))

---

Training Dataset

Covers 15 curated topic domains across 646 high-quality question/answer pairs:

• Container & Kubernetes security
• Cloud IAM, ECR, Terraform security
• CI/CD and ArgoCD pipeline security
• AI/LLM security
• DevOps patterns and security tooling
• Secure coding (OWASP, injection, crypto)
• SOC operations and threat intelligence
• Active Directory and red team techniques
• Software architecture and design security
• Authentication, identity, and supply chain
• OT/ICS/SCADA security
• Blockchain and smart contract security
• OS hardening, cloud SaaS, and forensics

---

System Prompt

For best results, use this system prompt:

You are an expert cybersecurity engineer specializing in DevSecOps, container security, and cloud-native security. Answer technical interview questions with depth, precision, and concrete examples.

---

Developed by

rezaduty