Skip to content

ThryLox/CogSec

Repository files navigation

🧠 Cognitive Threat Modeling Assistant (CTMA)

Human Factors Forensics for the Modern Cyber Attack Surface


🌐 Overview

The Cognitive Threat Modeling Assistant (CTMA) is a world-class forensic application that shifts the cybersecurity paradigm from "System-Centric" to "Human-Centric." It operates on the principle that the most critical vulnerability in any network is not a missing patch, but the human cognitive architecture.

By leveraging the Gemini 3 Pro engine, CTMA ingests high-resolution telemetry from the Cognitive Telemetry Sandbox (CTS) to reconstruct, analyze, and mitigate security failures caused by cognitive pressure, bias, and heuristics.


🛠️ System Architecture

graph LR
    A[Scenario Selection] --> B[CTS Telemetry Engine]
    B --> C{Counterfactual Engine}
    C --> D[Gemini Reasoning Pipeline]
    D --> E[Forensic Report Output]
Loading

1. Cognitive Telemetry Sandbox (CTS)

The CTS is the pulse of the application. It generates synchronized streams of data that model:

  • [C] Cognitive State: Real-time snapshots of workload level, time pressure, and attentional capacity.
  • [E] Environmental Events: System signals, alert arrival times, and visual similarity of stimuli.
  • [I] Interaction Logs: Micro-level user behavior, including decision latency (measured in milliseconds), warning dismissals, and click paths.

2. Counterfactual Reasoning Engine

A unique feature of CTMA that allows analysts to "rewind and modify" assumptions. By toggling variables like Alert Density or Urgency Cues, the AI calculates the probability of a different outcome, providing deep insights into which specific stressors were the "root cause" of a breach.


🔬 Scientific Foundations & Academic Citations

CTMA transforms abstract psychological concepts into actionable security intelligence:

📖 Dual Process Theory

  • Citation: Kahneman, D. (2011). Thinking, Fast and Slow.
  • Application: Analyzes the transition from analytical System 2 thinking to reactive, heuristic-based System 1 thinking under high-pressure scenarios like the Multitasking Interruption Storm.

📖 Cognitive Load Theory (CLT)

  • Citation: Sweller, J. (1988). Cognitive Load During Problem Solving.
  • Application: Quantifies the "Intrinsic Load" of security tasks. CTMA identifies the point of Cognitive Overload where working memory fails and security protocols are bypassed.

📖 Vigilance Decrement (The Mackworth Clock)

  • Citation: Mackworth, N. H. (1948). The Breakdown of Vigilance During Prolonged Visual Search.
  • Application: Modeled in the Alert-Fatigued Analyst scenario, tracking the decay of signal detection accuracy during long shifts.

📖 Automation Bias & Habituation

  • Citation: Parasuraman, R., & Manzey, D. H. (2010). Complacency and Bias in Human Use of Automation.
  • Application: Uses CTS data to detect habituation loops where users click "Dismiss" within <500ms due to visual similarity with previous benign alerts.

📑 The Forensic Pipeline (4 Stages)

When "Execute Forensic Analysis" is triggered, Gemini processes the CTS data through four rigorous stages:

  1. STAGE 1: Cognitive Reconstruction
    A temporal narrative of the user's mental journey. It maps stimuli arrival to shifts in perceived workload without judging the outcome.
  2. STAGE 2: Cognitive Vulnerability Inference
    The AI maps specific behavior (e.g., [I-2] latency of 0.4s) to active vulnerabilities like Automation Bias or Time Pressure Heuristics.
  3. STAGE 3: Counterfactual Reasoning
    The engine reasons about the "What If." If counterfactuals are enabled, it explicitly calculates the delta in risk if environmental stressors were reduced.
  4. STAGE 4: Human-Centered Mitigations
    Instead of proposing "better passwords," the system recommends Design-Level changes (e.g., variable alert delays to break habituation) and Training-Level interventions (e.g., metacognitive training).

⚖️ Ethical & Methodological Justification

The Use of Sandbox (CTS) Telemetry

CTMA exclusively uses Simulated Sandbox Telemetry for three critical reasons:

  1. Deterministic Replay: Forensic reasoning requires a "Ground Truth" that can be consistently re-evaluated under different counterfactual assumptions.
  2. Privacy-by-Design: Analyzing cognitive states is highly invasive. Using synthetic models allows for the study of patterns of failure without exposing real employees to "thought monitoring."
  3. Edge-Case Stress Testing: Real-world logs rarely capture the moment of "extreme overload." The sandbox allows us to simulate high-stress conditions that are necessary for robust threat modeling.

CTMA Core v2.1.4_STABLE | Engineered for the intersection of Mind and Machine.

About

An experimental Gemini-powered application that models human cognitive overload and bias as first-class cybersecurity attack surfaces, using simulated telemetry and multi-stage reasoning to generate explainable, human-centered security analyses.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors