🛰️ Orbital AI Infrastructure & Security Analysis Series

Complete framework analyzing AI transformation from terrestrial to orbital deployment

A systems-level analysis spanning physics, economics, security, and autonomous control from Earth to orbit.

---

📋 Table of Contents

• Overview
• The Series
• Key Findings
• Methodology
• Datasets
• Use Cases
• Technical Stack
• Related Work
• Connect
• Citation

---

🎯 Overview

This series provides a complete investment and engineering thesis for AI infrastructure transformation from Earth to orbit, analyzing five critical dimensions:

1. Physics & Economics (TX-1 Orbital - Why compute goes to orbit)
2. Terrestrial AI Security (OWASP LLM + RAG - Foundation threat models)
3. Orbital Threat Surface (2030 Threat Model - How security changes)
4. Control Constraints (Physics Over Humans - Why autonomy is required)
5. Economic Value Capture (Orbital Economics - Who wins)

Core Thesis:

By 2030, 40%+ of exascale AI training will operate from LEO/GEO. Traditional security models assume human intervention within decision windows. Orbital physics makes that impossible. Autonomous control becomes mandatory, not optional.

The question isn't "will this happen?"

The question is: "Who owns the orbital economy?"

Answer: Whoever guarantees continuity through autonomous control.

---

🚀 The Series

1. 🛰️ TX-1 Orbital Prototype: Physics & Economics of Compute in Space

Why orbital compute is inevitable

Key Metrics:

• Solar efficiency: 8× terrestrial (60MW sustained, free radiative cooling)
• Latency analysis: 2.7ms (LEO) to 240ms (GEO)
• Autonomy requirements: 63% local decision-making at high altitude
• Carbon offset: -0.9 tCO₂e per kWh index

Critical Findings:

Physics advantages:

• Free radiative cooling via Stefan-Boltzmann (σT⁴)
• 24/7 solar input (no atmospheric attenuation)
• Zero atmospheric drag
• No cooling infrastructure costs

Economic tipping point:

Terrestrial: $0.08-0.12/kWh + cooling + real estate
Orbital: $0.02-0.04/kWh (solar) + $0 cooling + launch amortized

When launch < $50/kg: Orbital becomes CHEAPER for high-density AI training

Timeline:

• 2025-2027: Prototypes (Starcloud H100 in orbit proves concept)
• 2028-2030: Commercial early adopters
• 2031-2033: Inflection point (5-10% of AI training)
• 2034-2037: Rapid shift (25-40% of exascale workloads)
• 2038+: Orbital dominance for large-scale training

Insight: Orbital compute isn't sci-fi. It's economics. When physics provides free cooling and constant solar, terrestrial data centers can't compete for power-hungry training.

📊 View Dashboard |

2. 🔐 OWASP LLM Attack Surface (2025 Edition)

Quantified risk analysis across the LLM stack

Key Metrics:

• Risk Ranking Matrix (Impact × Likelihood × Exposure Factor)
• Component Vulnerability Analysis (6 layers: Training → Agent)
• Attack Vector Prioritization (Kill Chain Analysis)

Critical Findings:

Threat	Risk Score	Systemic Impact
Prompt Injection	6.55	0.7820 (Direct)
Supply Chain Risk	5.52	0.5860 (Systemic)
Over-Agency	4.82	0.5230 (Amplification)

Top Vulnerability: Retrieval Layer

• 4 Critical + 3 High severity issues
• Single point of failure for RAG systems
• Propagates compromise downstream

Insight: Supply chain risk is systematically underestimated. Prompt injection affects ONE deployment. Supply chain compromise affects EVERY deployment using that component.

Not sexy. But systemic.

📊 View Dashboard |

---

3. 🎯 RAG Attack Surface Propagation Map (2025 Edition)

Cascading failure analysis across RAG pipelines

Key Metrics:

• Propagation Risk Formula: (Source_Vulnerability × Coupling_Factor) × Downstream_Exposure
• Weak-Link Index: Identifies critical failure pairs
• Attack Path Matrix: How single weakness cascades

Critical Findings:

Weakest Link Pair:

• Embedding Model (0.815) + Vector Database (0.792) = 0.79 combined vulnerability
• If EITHER compromised → downstream contamination propagates

Highest Risk Attack:

• Prompt Injection: 6.55 risk score (2× higher than next threat)
• Tool Misuse: 3.28 (exponential risk in tool-using RAG)

Propagation Example:

Embedding poisoned (0.88) → 
Vector DB corrupted (0.92) → 
Retrieval manipulated (0.80) → 
LLM outputs toxic (0.75)

Insight: RAG orchestrators (0.7620 risk) are invisible to most monitoring but silently amplify attacks. Most teams secure the LLM. The vulnerability is in the retrieval layer.

📊 View Dashboard |

---

4. 🛰️ Orbital AI Security 2030: Threat Model

When humans leave the loop, AI systems fail differently

Key Metrics:

• Threat Surface Shift (Earth vs Orbital risk severity)
• Control Stress Analysis (environmental strain on security models)
• Failure Condition Triggers (what breaks first)

Critical Findings:

Threat Surface Comparison (Earth → Orbital):

Factor	Earth	Orbital	Delta
Autonomy Requirement	1.6	4.7	+194%
Incident Response Window	2.2	4.5	+105%
Physical Access Risk	1.8	4.2	+133%
Patch Latency	2.1	3.8	+81%

What Breaks FIRST in Orbit:

1. Autonomy Control (4.70 risk) - Humans can't respond within 50ms decision windows
2. Incident Response (4.50 risk) - Response time > latency = failure
3. Physical Access (4.20 risk) - Can't drive a truck to LEO with a USB stick

Insight: The dominant threat in space AI is not intrusion—it's loss of control.

Traditional security assumes human-in-the-loop. Orbital physics makes that impossible.

📊 View Dashboard |

---

5. ⚡ Physics Over Humans: The 2030 Control Shift

Quantifying when security models fail under latency constraints

Key Metrics:

• Control Strain Analysis (how much stress each model experiences)
• Latency Tolerance Thresholds (when models collapse)
• Monte Carlo Risk Impact (100,000 simulations per attack vector)

Critical Findings:

Control Strain Under Latency:

Security Model	Latency Tolerance	Orbital Strain	Status
Human-in-the-Loop	4,500ms	0.91	CRITICAL
Incident Response	1,800ms	0.93	CRITICAL
Zero Trust	1,200ms	0.88	CRITICAL
Autonomous Control	120ms	0.48	STABLE

Risk Impact (Monte Carlo Simulation - 100k runs):

Attack Vector	Human-in-Loop	Incident Response	Zero Trust	Autonomous
Data Poisoning	10.2	9.3	8.6	9.0
Model Drift	10.5	9.0	8.8	9.2
Prompt Injection	10.1	9.2	8.5	8.9
Tool Misuse	10.3	9.1	8.7	9.1
Unauthorized Actions	10.7	9.8	9.0	9.4

Overall Risk Averages:

• Human-in-the-Loop: 10.9 (highest - fails across all vectors)
• Incident Response: 9.6 (unstable)
• Zero Trust: 8.7 (stressed but surviving)
• Autonomous Control: 9.1 (only latency-native model)

Insight: Human-in-the-Loop consistently performs WORST across every attack vector. At >4.5s latency, human decision cycles are no longer viable for security enforcement.

The Uncomfortable Truth:

Architecture doesn't fail because engineers don't try hard enough.

It fails because reality eventually enforces its constraints.

📊 View Dashboard |

---

6. 💰 Orbital Economics: Markets Only Exist After Continuity is Guaranteed

Constraint-driven economics of autonomous orbital systems

Key Metrics:

• Orbital Value Stack (who captures value at each layer)
• Risk Concentration Analysis (launch vs control risk)
• Value Capture vs Failure Absorption (payoff diagram)

Critical Findings:

The Orbital Value Stack:

Layer	Value Capture	Risk Profile
Autonomous Control	92	Infinite (every decision, forever)
Networking/Downlink	60	Moderate (visible but commoditizing)
Compute	55	Depreciating asset
Storage	48	Commodity
Power/Energy	42	Free input (after capital cost)
Launch	35	Finite (high upfront, then done)

The Inversion:

Traditional Cloud: Compute = valuable, control = commodity
Orbital Cloud: Control = valuable, compute = commodity

As you move UP the orbital stack:

• Capital intensity DECREASES
• Risk concentration INCREASES
• Value capture INCREASES

Launch Risk vs. Control Risk:

Launch:

• High upfront (rocket could explode)
• But FINITE (once in orbit, done)
• Risk score: 55

Control:

• Lower per-decision
• But INFINITE (every decision, forever)
• Risk score: 92

Control risk compounds. Launch risk doesn't.

Core Thesis: Markets only exist after continuity is guaranteed.

Insight:

Whoever can GUARANTEE continuity despite:

• 240ms latency
• Disconnection events
• Physical inaccessibility
• Adversarial attacks

Owns the orbital economy.

A loss of control cascades across compute, networking, storage, and power—destroying economic value faster than any physical component failure could.

In orbit, control risk is existential risk.

Control = single point of economic fragility.

📊 View Dashboard |

---

🔥 Key Findings Across Series

Finding 1: Orbital Compute Is Inevitable (TX-1 Analysis)

Timeline:

• 2031-2033: Inflection point (5-10% of AI training)
• 2034-2037: Rapid shift (25-40% of exascale workloads)
• 2038+: Orbital dominance

Why:

• Solar: 8× terrestrial efficiency, 24/7 availability
• Cooling: Free radiative (σT⁴), zero infrastructure cost
• Economics: When launch < $50/kg, orbital < terrestrial cost

What happens to terrestrial data centers:

• New construction slows 15-20% (2032-2035)
• Facilities pivot: inference + edge + compliance workloads
• Ground control becomes new value (data staging, mission control)

Texas doesn't die. It pivots.

---

Finding 2: Supply Chain Risk is Systematically Underestimated (OWASP)

From OWASP Analysis:

• Supply Chain: 0.5860 systemic impact
• Prompt Injection: 0.7820 direct impact

Why it matters:

• Prompt injection affects ONE deployment
• Supply chain compromise affects EVERY deployment

Real-world parallel:

• xz backdoor (2024): Almost shipped in every Linux distro
• If successful: SSH access to millions of servers
• AI equivalent: Poisoned model weights, compromised frameworks

Not sexy. But systemic.

---

Finding 3: RAG Systems Have a Critical Weak Link (RAG Propagation)

From RAG Propagation Analysis:

• Embedding Model + Vector Database = 0.79 combined vulnerability
• If EITHER compromised → cascading failure

Attack propagation:

Single weakness → 
Embedding poisoned (0.88) → 
Vector DB corrupted (0.92) → 
Retrieval manipulated (0.80) → 
Outputs toxic (0.75)

Insight: Most teams secure the LLM. The vulnerability is in the retrieval layer.

RAG orchestrators (0.7620) are invisible to monitoring but silently amplify attacks.

---

Finding 4: Orbital Systems Require Autonomous Security (2030 Threat)

From 2030 Threat Model:

• Autonomy requirement: 4.7× higher in orbit vs Earth
• Incident response: 4.5× tighter windows
• Physical access: 4.2× higher risk (can't patch physically)

What breaks first:

1. Autonomy control (4.70 risk)
2. Incident response (4.50 risk)
3. Physical access (4.20 risk)

Traditional security = perimeter defense. Orbital security = control continuity.

---

Finding 5: Human-in-the-Loop Fails Under Latency (Physics Over Humans)

From Physics Over Humans Analysis:

Monte Carlo Simulation (100k runs):

• Human-in-the-Loop: 10.9 average risk (worst across ALL vectors)
• Autonomous Control: 9.1 average risk (only viable)

Control strain:

• Human-in-the-Loop: 0.91 orbital strain (critical failure)
• Autonomous Control: 0.48 orbital strain (stable)

The math is brutal:

Decision window: 50ms
Orbital latency: 240ms (GEO)
Human response time: 4,500ms

When response > latency:
→ Humans can't intervene
→ Controls degrade
→ Systems fail

Security failure is architectural, not operational.

---

Finding 6: Value Concentrates at Autonomous Control Layer (Orbital Economics)

From Orbital Economics Analysis:

• Autonomous Control: 92 value capture (dominates)
• Launch: 35 value capture (commoditizes)

The inversion:

Traditional cloud: Compute valuable, control commodity
Orbital cloud: Control valuable, compute commodity

Why?

• Launch risk = finite (once in orbit, done)
• Control risk = infinite (every decision, forever)
• Control failures cascade across ALL infrastructure layers
• Economic value destroyed faster by control loss than hardware failure

The thesis:

Markets only exist after continuity is guaranteed.

Whoever can guarantee continuity through autonomous control = owns the orbital economy.

This isn't just technology analysis. It's investment thesis.

---

📊 Methodology

Risk Quantification Framework

OWASP LLM & RAG Analysis:

Risk = Impact × Likelihood × Exposure Factor

Where:
- Impact: Severity of successful exploit (1-10 scale)
- Likelihood: Probability of attack succeeding (0-1)
- Exposure Factor: Systemic vs. direct impact multiplier

Priority Index:

Priority = (Risk_Score × Component_Criticality) / Mitigation_Effort

---

Control Strain Analysis

Physics Over Humans & 2030 Threat Model:

Control_Strain = (Latency_Constraint / Model_Tolerance) × Environment_Factor

Where:
- Latency_Constraint: Round-trip communication delay
- Model_Tolerance: Maximum latency model can handle
- Environment_Factor: Multiplier based on connectivity/reliability

When Control_Strain > 0.8 → Critical failure zone
When Control_Strain > 0.9 → Model collapse imminent

Example (Human-in-the-Loop in Orbital):

Latency: 240ms (GEO round-trip)
Human Response: 4,500ms (decision + authorization)
Total Constraint: 4,740ms
Model Tolerance: 4,500ms

Control_Strain = 4,740 / 4,500 × 0.95 = 0.91 (CRITICAL)

---

Propagation Risk Modeling

RAG Attack Surface:

Propagation_Risk = (Source_Vulnerability × Coupling_Factor) × Downstream_Exposure

Weak_Link_Index = min(Component_A, Component_B) where coupling > 0.7

---

Economic Value Modeling

Orbital Economics:

Value_Capture = (Control_Authority × Risk_Absorption) / Capital_Intensity

Risk_Concentration = Σ(Downstream_Failures) × Cascade_Factor

Launch_Risk = High × Finite
Control_Risk = Moderate × Infinite

Control_Risk > Launch_Risk (over time)

---

Monte Carlo Simulation

Physics Over Humans:

• 100,000 simulations per attack vector/model combination
• Mean risk impact scored on 10-point scale
• Pivoted analysis across all failure modes
• Statistical validation: Patterns repeatable and predictable

---

📁 Datasets

All datasets used in this analysis are available in each project repository:

TX-1 Orbital Prototype

• orbital_latency_analysis.csv - LEO/GEO latency measurements
• solar_efficiency_calculations.csv - Power generation vs terrestrial
• autonomy_requirements.csv - Decision-making thresholds by altitude
• economic_tipping_points.csv - Cost comparisons terrestrial vs orbital

OWASP LLM Attack Surface

• owasp_llm_risk_matrix.csv - Risk scores by threat vector
• component_vulnerability_scores.csv - Infrastructure layer analysis
• attack_kill_chain.csv - Propagation paths
• priority_index_calculations.csv - Mitigation prioritization

RAG Propagation Map

• rag_component_vulnerabilities.csv - Individual component scores
• propagation_matrix.csv - Cascading failure paths
• weak_link_analysis.csv - Critical pair identification
• attack_vector_impact.csv - Risk scores by attack type

Orbital AI Security 2030

• threat_surface_earth_vs_orbital.csv - Comparative risk analysis
• control_stress_by_environment.csv - Strain calculations
• failure_condition_triggers.csv - What breaks first analysis
• timeline_projections.csv - 2025-2038+ deployment estimates

Physics Over Humans

• control_strain_analysis.csv - Latency vs model tolerance
• monte_carlo_risk_impacts.csv - 100k simulation results
• latency_tolerance_thresholds.csv - Breaking points by model
• attack_vector_simulation_results.csv - Risk by failure mode

Orbital Economics

• orbital_value_stack.csv - Value capture by layer
• risk_concentration_analysis.csv - Launch vs control risk
• value_capture_vs_failure_absorption.csv - Payoff matrix
• economic_timeline_projections.csv - Market evolution 2025-2040

Data Format: CSV, ready for analysis in Python/R/Excel

License: MIT (free to use with attribution)

---

🎓 Use Cases

For Security Architects:

• Design autonomous security frameworks for high-latency environments
• Understand terrestrial → orbital threat surface transformation
• Prioritize vulnerabilities by systemic impact (not just direct risk)
• Plan transition from reactive to autonomous security

For MLOps Engineers:

• Secure RAG pipelines (identify weak-link vulnerabilities)
• Model drift detection strategies
• Supply chain security for ML dependencies
• Production deployment threat modeling

For Space Systems Engineers:

• Model latency constraints for orbital AI deployments
• Design fail-autonomous architectures
• Plan redundancy strategies (N+2 minimum)
• Autonomous control frameworks

For VCs & Investors:

• Understand economic value concentration (who wins in orbital compute)
• Evaluate orbital infrastructure startups (control layer vs hardware layer)
• Assess market timing (2030-2035 inflection point)
• Investment thesis: Autonomous control = 92 value capture

For AI Safety Researchers:

• Study failure modes when human oversight is physically impossible
• Design pre-authorized decision frameworks
• Model risk distribution across autonomous systems
• Validate Monte Carlo threat modeling approaches

For Policy Makers:

• Understand regulatory implications of autonomous AI in space
• Define liability frameworks when humans can't intervene
• Balance safety requirements with physical constraints
• Plan for 2030+ orbital AI governance

For CTOs / VPs Engineering:

• Assess organizational MLOps maturity
• Quantify AI security investment priorities
• Understand future infrastructure shifts (terrestrial → orbital)
• Build teams for autonomous AI operations
• Evaluate "build vs buy" for orbital compute (2030-2035 timeframe)

---

🧰 Technical Stack

Analysis & Modeling:

• Python (pandas, NumPy, SciPy)
• Monte Carlo simulation framework
• Statistical modeling (risk quantification)
• Physics-based constraint modeling
• Economic value modeling

Visualization:

• Tableau Public (interactive dashboards)
• Matplotlib/Seaborn (exploratory analysis)
• Custom risk heatmaps
• Bubble charts (value capture vs failure absorption)

Data Engineering:

• CSV pipelines (reproducibility)
• Version control (Git/GitHub)
• Documentation (Markdown)

Frameworks Applied:

• OWASP LLM Top 10 (2025)
• MITRE ATT&CK for ML
• CAP theorem (distributed systems)
• Control theory (latency constraints)
• Economic value chain analysis

---

📚 Related Work

Companion Projects:

Texas Energy & Infrastructure:

• 🔋 Texas Energy Data Pulse - 20 dashboards analyzing AI infrastructure demands
• 🛰️ TX-1 Orbital Prototype - Physics and economics of compute in space (this series)

AI Security Foundations:

• 🔐 OWASP LLM Attack Surface (this series)
• 🎯 RAG Propagation Map (this series)

Orbital Security Framework:

• 🛰️ Orbital AI Security 2030 (this series)
• ⚡ Physics Over Humans (this series)
• 💰 Orbital Economics (this series)

Production MLOps:

• 🤖 MLOps Maturity Framework 2026 (coming soon)
• 🐧 Linux Infrastructure for AI (coming soon)

Together, these projects form a complete framework:

Earth infrastructure → Orbital compute → Security transformation → Autonomous operations → Economic value capture

---

🔗 Connect

Tracy Manning
Staff MLOps Engineer | Multi-Cloud + Linux Infrastructure Architect | Austin, TX

💼 LinkedIn
🐦 X/Twitter
📊 Tableau Public
📱 WhatsApp Channel

Building AI systems that respect physics, not wishful thinking.

---

📖 Citation

If you reference this work:

@misc{manning2025orbitalaiseries,
  author = {Manning, Tracy},
  title = {Orbital AI Infrastructure \& Security Analysis Series: Complete Framework from Terrestrial to Orbital Deployment},
  year = {2025},
  publisher = {GitHub},
  howpublished = {\url{https://github.com/TAM-DS/Orbital-AI-Security-Series}},
  note = {Comprehensive framework analyzing AI infrastructure transformation from Earth to orbit, including physics, economics, security threat models, control constraints, and economic value capture. Includes TX-1 Orbital Prototype, OWASP LLM attack surfaces, RAG propagation modeling, 2030 threat analysis, physics-constrained control systems, and orbital economics analysis.}
}

---

📄 License

MIT License - Free to use with attribution.

See individual project repositories for specific licensing details.

---

🙏 Acknowledgments

• Security researchers validating OWASP LLM Top 10 framework
• Space systems engineers reviewing orbital latency constraints
• AI safety community exploring autonomous decision frameworks
• MLOps practitioners sharing production failure modes
• Economics researchers analyzing value chain transformations
• VCs and investors providing market validation
• Open source community enabling reproducible research

---

🎯 Key Takeaway

By 2030, 40% of exascale compute will operate from LEO/GEO.

Traditional security models assume human intervention.

Orbital physics makes that assumption obsolete.

The question isn't "should we build autonomous AI security?"

The question is: "Who owns the orbital economy?"

Answer: Whoever guarantees continuity through autonomous control.

Markets only exist after continuity is guaranteed.

This series provides the complete framework:

• WHY orbital compute happens (physics + economics)
• HOW security transforms (threat models + control constraints)
• WHO WINS (autonomous control = 92 value capture)

Not just analysis. Investment thesis.

---

"Architecture doesn't fail because engineers don't try hard enough. It fails because reality eventually enforces its constraints."

"Markets only exist after continuity is guaranteed."

---

⭐ Star this repo if you find it useful!

🔔 Watch for updates as the series expands!

💬 Questions? Open an issue or connect on LinkedIn!

---

📊 Complete Dashboard Series:

1. TX-1 Orbital Prototype (Physics & Economics)
2. OWASP LLM Attack Surface (Terrestrial Security)
3. RAG Propagation Map (Cascading Failures)
4. Orbital AI Security 2030 (Threat Surface Shift)
5. Physics Over Humans (Control Constraints)
6. Orbital Economics (Value Capture)

= Six elite dashboards. One complete thesis. Earth → Orbit → Security → Economics.