Cognitive Security & Information Operations — Disinformation, Influence & Cognitive Hacking

Ringkasan & Hubungan ke Vault

Cognitive security adalah Layer 8 (Human) dari keamanan — bukan exploit teknis, tapi exploit psikologis. Catatan ini melengkapi osint dengan dimensi influence operations, dan digital-privacy-anonymity dengan ancaman cognitive hacking. Relevan dengan era AI-generated content yang membuat disinformation scalable.

Domain: Cyber Security / Human Factors Tags: cognitive-security information-operations disinformation deepfake influence

1. Cognitive Security Framework

1.1 The OODA Loop for Information Warfare

Observe → Orient → Decide → Act
   ↑                          |
   └──────────────────────────┘

Attacker manipulates:
├── Observe: filter bubbles, selective exposure
├── Orient: framing, narrative control, false context
├── Decide: cognitive biases manipulation
└── Act: social proof, manufactured consent

1.2 Attack Surface

Vector	Method	Example
Social Media	Bot farms, astroturfing	2016 US election
Deepfake	AI-generated video/voice	Fake CEO voice call ($243K heist)
Content Farm	SEO manipulation	Misinformation sites rank high
Weaponized Narrative	Frame control	False flag narratives
Astroturfing	Fake grassroots support	Manufactured consensus
Cognitive Hacking	Memory manipulation	Gaslighting at scale
AI-Generated Text	LLM propaganda	ChatGPT-generated disinformation

2. Deepfake Detection

2.1 Detection Techniques

# Deepfake detection — frequency domain artifacts
import cv2
import numpy as np
 
def detect_deepfake(frame):
    # 1. Check facial blending artifacts
    face = extract_face(frame)
 
    # 2. Frequency analysis — GAN faces lack high-freq detail
    fft = np.fft.fft2(cv2.cvtColor(face, cv2.COLOR_BGR2GRAY))
    fft_shift = np.fft.fftshift(fft)
    magnitude = np.log(np.abs(fft_shift) + 1)
 
    # 3. Check eye reflection consistency
    left_eye, right_eye = extract_eyes(face)
    if not reflections_match(left_eye, right_eye):
        return "DEEPFAKE: inconsistent eye reflections"
 
    # 4. Check blinking rate (old deepfakes lacked blinks)
    blink_rate = measure_blink_rate(video_sequence)
    if blink_rate < threshold:
        return "DEEPFAKE: abnormal blink pattern"
 
    return "AUTHENTIC"

3. Koneksi ke Vault

Note	Hubungan
osint	OSINT countermeasures, source verification
digital-privacy-anonymity	Privacy as defense against profiling
llm-security-red-teaming-attack-surface-ai-layer	LLM-generated disinformation detection
social-engineering	Extension to population-scale manipulation

---

📚 Referensi

1. “The Hacker and the State” — Ben Buchanan
2. “LikeWar” — Singer, Brooking
3. Deepfake detection: https://deepfakedetection.ai/
4. NATO STRATCOM COE reports