Press any key or click to skip

Back to projects

CogSec — Cognitive Threat Modeling

Gemini APIAI/MLPythonSecurity Research
Source

Overview

CogSec shifts cybersecurity from system-centric to human-centric analysis. Using Gemini 3 Pro, it reconstructs and analyzes security failures caused by cognitive pressure, bias, and decision fatigue.

The system treats the human operator as the primary vulnerability vector—not the missing patch.

How It Works

  1. Telemetry Capture — Sandbox collects response times, click patterns, and cognitive load indicators
  2. Vulnerability Inference — Maps behaviors to cognitive biases (automation bias, time pressure heuristics)
  3. Counterfactual Reasoning — Calculates "what if" risk deltas under modified conditions
  4. Human-Centered Mitigations — Recommends design changes, not just technical fixes

Key Innovation

Traditional security focuses on technical controls. CogSec asks: "What cognitive state led to this failure?"

Outputs include forensic timelines, vulnerability mappings, and actionable interventions targeting alert fatigue, habituation, and decision overload.

Tech Stack

Gemini API Python Cognitive Psychology Forensic Analysis