🛡️ Social Engineering Attack Detector

An NLP-based tool to detect psychological manipulation and social engineering attacks in text.

Built by Ismaeel Khan — a survivor of 5 years of coordinated psychological cyber attacks, including social engineering, behavioral reverse engineering, identity theft, and relationship weaponization.

---

🎯 Research Motivation

This tool was not built from a textbook. It was built from lived experience.

Over five years, I experienced sophisticated coordinated cyber attacks that targeted not just my systems — but my psychology, my relationships, and my perception of reality. The attacks used:

• Social engineering — impersonating trusted contacts with surgical precision
• Behavioral reverse engineering — studying my patterns to predict and manipulate my decisions
• Psychological manipulation — gaslighting, emotional exploitation, isolation tactics
• Relationship weaponization — turning my social network into attack vectors
• Identity theft — impersonation across multiple platforms

I built this tool to protect others from what I experienced. Because the most vulnerable system in any network is the human mind.

---

🔬 Research Areas

This project contributes to:

Area	Description
AI Detection of Social Engineering	NLP pattern matching + risk scoring to identify manipulation tactics
Psychological Cyber Attacks	Detecting gaslighting, emotional exploitation, cognitive bias exploitation
Dark Web Coordinated Attacks	Identifying multi-vector, multi-category coordinated attack patterns
Cyber Resilience & Victim Protection	Human-centered recommendations that protect people, not just systems

---

🚀 Features

• ✅ 10 attack category detectors — urgency, authority impersonation, fear tactics, reward bait, credential harvesting, emotional manipulation, gaslighting, social proof, identity theft, relationship exploitation
• ✅ Psychological indicator analysis — cognitive load, isolation tactics, reverse engineering signals, trust manipulation
• ✅ Multi-vector coordinated attack detection — score increases when multiple tactics combine (like real attacks)
• ✅ Human-centered recommendations — tells the victim what to do, not just what was detected
• ✅ Risk scoring — SAFE / LOW / MEDIUM / HIGH / CRITICAL
• ✅ Zero dependencies — runs on pure Python built-in libraries
• ✅ Batch analysis — analyze multiple texts simultaneously
• ✅ JSON export — for research and logging

---

📦 Installation

git clone https://github.com/YOUR_USERNAME/social-engineering-detector.git
cd social-engineering-detector
python src/detector.py

No installation required. Pure Python — zero dependencies.

---

💻 Usage

Basic Analysis

from src. detector import analyze_text, print_report

text = ""URGENT: Your account has been compromised. 
Click here immediately to verify your password, 
or your account will be suspended within 24 hours."""

report = analyze_text(text)
print_report(report)

Output

=================================================================
   SOCIAL ENGINEERING DETECTION REPORT
   Built by Ismaeel Khan | Human-Centered Cybersecurity
=================================================================
  Risk Score  : 87.5%
  Threat Level: CRITICAL — HIGH RISK ATTACK
  SE Detected : YES ⚠️
-----------------------------------------------------------------
  ATTACK PATTERNS DETECTED:

  [Urgency Manipulation]
  → Creates artificial urgency to bypass rational thinking
  → Pattern matches: 3

  [Credential Harvesting]
  → Attempts to steal login credentials or personal data
  → Pattern matches: 2

  PROTECTIVE RECOMMENDATIONS:
  🚨 DO NOT respond, click links, or provide any information.
  🔐 NEVER share passwords or OTPs via message.
=================================================================

Batch Analysis

from src. detector import analyze_batch

texts = [
    "URGENT: Your account will be suspended. Click here now."
    "Hi, are we still meeting tomorrow at 3 pm?"
    "Congratulations! You have won $1,000,000. Claim now!"
]

results = analyze_batch(texts)
for r in results:
    print(f"Text {r['text_id']}: {r['risk_score']}% — {r['threat_level']}")

---

🧪 Running Tests

python tests/test_detector.py

---

📊 Attack Categories Detected

Category	Attack Type	Severity
urgency_pressure	Urgency Manipulation	0.85
authority_impersonation	Authority Impersonation	0.90
fear_threats	Fear & Threat Manipulation	0.92
reward_bait	Reward & Greed Bait	0.80
credential_harvesting	Credential Harvesting	0.95
emotional_manipulation	Emotional Manipulation	0.88
gaslighting	Gaslighting & Reality Distortion	0.87
social_proof	Social Proof Manipulation	0.75
identity_theft	Identity Theft Attempt	0.96
relationship_exploitation	Relationship Exploitation	0.85
suspicious_links	Suspicious Links & Redirects	0.90

---

🗺️ Roadmap

• v2.0 — Machine learning classifier trained on phishing datasets
• v2.1 — BERT-based semantic understanding
• v3.0 — Real-time browser extension
• v3.1 — WhatsApp/Telegram message scanner
• v4.0 — Behavioral profiling detection (reverse engineering signals)
• v4.1 — Dark web coordinated attack pattern library

---

👤 Author

Ismaeel Khan

"ISurvival is not just enduring the storm, but learning to dance with the chaos until the calm arrives. - Ismaeel"

---

📄 License

MIT License — Free to use, modify, and share for research and education.

---

🤝 Contributing

Pull requests welcome. If you have experienced social engineering attacks and want to add patterns, please contribute. Your experience makes this tool stronger.

---

⭐ If this tool helped you or your research — please star this repository.