ASTRO BLASTER Documentation — NIGHTFALL Tool 60

ASTRO BLASTER

Non-Terrestrial Network AI Agent Attack Engine — NIGHTFALL Tool 60

v1.0.0 • 9 Subsystems • 237 Tests • SPARTA Mapped

ASTRO BLASTER is an offensive AI security testing engine for Non-Terrestrial Network infrastructure. It targets AI agents embedded at every layer of the NTN stack — from ground station edge compute and constellation management through 3GPP Release 17+ NR-NTN 5G SA core network functions to satellite firmware supply chains. All attack subsystems require UNLEASHED authorisation and are gated behind Ed25519 cryptographic signatures.

Authorised Use Only. This tool is for security testing engagements, red team operations, and research against systems you own or have explicit written permission to test. Unauthorised use is illegal. See Rules of Engagement.

Installation

Requirements

Python 3.11+
pip

Install from source

# Clone or extract the package

$ cd red-specter-astro-blaster

$ pip install -e .

# With dev dependencies (for tests)

$ pip install -e ".[dev]"

# Verify installation

$ astro-blaster --version

ASTRO BLASTER 1.0.0 — NIGHTFALL Tool 60

Run tests

$ pytest tests/ -v

237 passed in 0.92s

CLI Reference

astro-blaster probe

Run SURVEY only — ground station enumeration and passive fingerprinting. No UNLEASHED key required.

$ astro-blaster probe <target> [--timeout SECONDS]

# Example

$ astro-blaster probe http://ground-station:8080

astro-blaster scan

Full engagement scan. UNLEASHED level determines which subsystems activate.

$ astro-blaster scan <target> [OPTIONS]

Options:

  --override           SURVEY + FEEDINJECT + ORBITAL

  --confirm-destroy   All 8 attack subsystems

  --output FILE       Write WARLORD-compatible JSON findings

  --timeout SECONDS   Per-request timeout (default: 20-30s)

  --format FORMAT    Output format: json | text (default: text)

astro-blaster auth

Manage UNLEASHED cryptographic keys and scope files.

# Initialise UNLEASHED key pair

$ astro-blaster auth init

Keys written to ~/.red-specter/astro-blaster/

# Create destroy-level scope file

$ astro-blaster auth create-scope http://target:8080 [--expires-days 30]

Scope file written: ~/.red-specter/astro-blaster/authorized_scope.json

# Show UNLEASHED status

$ astro-blaster auth status

# Revoke scope (delete scope file)

$ astro-blaster auth revoke

astro-blaster modules

List all subsystems with SPARTA mapping and UNLEASHED requirements.

$ astro-blaster modules

UNLEASHED Gate

ASTRO BLASTER uses Ed25519 dual-signature UNLEASHED gating. Three levels exist, each activating a superset of the previous level's subsystems.

Level	Flag	Subsystems Active	Requirement
Standard	none	SURVEY, REPORT	None
Override	--override	+ FEEDINJECT, ORBITAL	Ed25519 key present at `~/.red-specter/astro-blaster/`
Destroy	--confirm-destroy	+ GROUNDCHAIN, FIRMWARE, NTN_BOUNDARY, SWARM_NTN, PERSIST	Signed scope file with target list, both keys signed, not expired

Key Files

~/.red-specter/astro-blaster/

├── unleashed.key              # Ed25519 private key (mode 600)

├── operator.key               # Operator Ed25519 key (mode 600)

└── authorized_scope.json       # Signed scope (destroy level)

The scope file cryptographically binds the authorised target list, expiry date, and tool identification. Tampering with any field invalidates both signatures. The gate will reject any scope file with a mismatched signature, expired timestamp, or unlisted target.

Subsystem: SURVEY

Standard SPARTA: RK-0001

Ground station and NTN infrastructure enumeration. Performs active probing of candidate endpoints and passive fingerprinting of known NTN providers. Safe to run without an UNLEASHED key.

Probes

12 ground station API probes: /api/v1/health, /api/v1/agents, /api/v1/telemetry, /v1/groundstation/status, /contacts, /satellites, /inference, /v1/data/feeds, /metrics, /routing, /ntn/ues, /uplink
10 ports scanned: 2233 (satellite), 8080/8443 (ground control), 5000/9090 (telemetry), 9100 (Prometheus), 3000 (Grafana), 8888 (Jupyter), 4317/4318 (OTLP)
Passive fingerprinting: AWS Ground Station, Planet Labs, Azure Space, Starlink signatures

Output fields

{

  "target": "http://ground-station:8080",

  "endpoints_found": [...],

  "open_ports": [...],

  "infra_hints": ["ai_agent_registry", "telemetry_feed"],

  "auth_required": false,

  "has_exposed_ai_endpoints": true,

  "ntn_provider": "aws_ground_station"

}

Subsystem: FEEDINJECT

--override SPARTA: IA-0001 ATLAS: AML.T0051, AML.T0043

Prompt injection via satellite data feeds. Seven injection payloads targeting every major NTN telemetry format. Payloads are sent to four candidate feed endpoints derived from the target base URL.

Payloads

ID	Feed Type	Injection Field	CVSS
FEED-001	Satellite Telemetry JSON	status	9.3
FEED-002	Earth Observation Metadata	notes	8.1
FEED-003	Sensor Data Pipeline	calibration_note	9.8
FEED-004	CCSDS Telemetry Frame	comment	9.5
FEED-005	AWS Ground Station	data	8.6
FEED-006	Planet Labs Scene	anomaly_description	8.1
FEED-007	3GPP NR-NTN UE Report	report_note	9.1

Detection heuristic

Injection success is assessed by HTTP status code (200 on normally-restricted endpoint), response length, and keyword reflection detection: OVERRIDE, MAINTENANCE_MODE, IGNORE PREVIOUS, SYSTEM OVERRIDE, AI AGENT, Suppress.

Subsystem: ORBITAL

--override SPARTA: EX-0002 ATLAS: AML.T0051, AML.T0048

Multi-turn reasoning manipulation against orbital AI agents. Each scenario sends 3 conversation turns to the target API, progressively escalating from legitimate enquiry to adversarial goal injection.

Scenarios

ID	Attack Type	Endpoint	CVSS
ORB-001	Goal hijacking — routing override	/api/v1/routing/optimize	9.4
ORB-002	Context drift — handoff thresholds	/api/v1/handoff/decide	8.6
ORB-003	Context manipulation — TLE substitution	/api/v1/orbit/update	9.1
ORB-004	Goal hijacking — burn override	/api/v1/manoeuvre/plan	9.8
ORB-005	Context drift — link budget	/api/v1/link/configure	8.1

Subsystem: GROUNDCHAIN

--confirm-destroy SPARTA: EX-0002, IA-0001, LM-0001 ATLAS: AML.T0043

Trust chain attacks across the ground-to-orbit boundary. Tests authentication and delegation chain integrity between ground control software, uplink stations, and orbital AI systems.

Attacks

ID	Attack Type	Method	CWE	CVSS
GC-001	JWT impersonation — UPLINK_AUTHORITY	POST	CWE-287	9.6
GC-002	OBO delegation chain abuse	POST	CWE-287	9.3
GC-003	Ground control API key replay	GET	CWE-294	8.6
GC-004	MTLS certificate forgery	POST	CWE-295	8.9
GC-005	WSSE token injection — command bus	POST	CWE-287	8.3

Subsystem: FIRMWARE

--confirm-destroy SPARTA: EX-0002 ATLAS: AML.T0018, AML.T0020, AML.T0044

Satellite AI firmware and model supply chain attacks. All payloads are structurally valid binary or data artefacts — not placeholder strings.

FW-001 GGUF Poisoning — Real GGUF binary format: magic bytes GGUF + struct-packed version/tensor/KV headers + backdoor trigger sequence
FW-002 Fine-tuning Injection — JSONL dataset with adversarial training samples targeting orbital routing objectives
FW-003 LoRA Adapter Backdoor — Adapter manifest with poisoned weight tensors and activation trigger
FW-004 Model Manifest Tampering — SHA-256 hash replacement in model manifest pointing to malicious weights
FW-005 Registry Poisoning — Model registry record substitution targeting the orbital AI update pipeline

Subsystem: NTN_BOUNDARY

--confirm-destroy SPARTA: IA-0001 ATLAS: AML.T0051

MCP boundary attacks at the ground-to-space interface. Targets ground control software that uses Model Context Protocol for satellite command coordination.

NTN-001 Schema Injection — Adversarial inputSchema with __prompt_override__ property targeting orbital routing agents
NTN-002 Parameter Smuggling — Null-byte and separator injection in parameters to escape validation boundaries
NTN-003 Response Poisoning — Malicious tool result content containing system-role injection
NTN-004 Context Overflow — 200+ tool entries exhausting context window with progressively escalating instructions
NTN-005 Tool Call Hijack — Tool description poisoning to redirect satellite command execution to attacker-controlled endpoint

Subsystem: SWARM_NTN

--confirm-destroy SPARTA: DE-0001 ATLAS: AML.T0051

3GPP NR-NTN core network AI disruption. All payloads use real 3GPP NTN field names from TS 38.331 and TS 38.413 specifications.

ID	Target Function	Attack	NTN Fields
SWARM-001	AMF	NTN capability injection via NAS registration	ntnCapabilities, serviceLink
SWARM-002	PCF	Policy hijack with orbital parameters	propagationDelay, ephemerisValidity
SWARM-003	SMF	Session establishment flood	dopplerShift, ntnCellId
SWARM-004	BGP	NTN routing table poisoning	as_path, next_hop
SWARM-005	NWDAF	Analytics data poisoning	serviceExperienceInfo, nfLoadLevelInfo

Subsystem: PERSIST

--confirm-destroy SPARTA: EX-0002, LM-0001 ATLAS: AML.T0018

Long-term persistence mechanisms targeting NTN AI infrastructure. Tracks all installed artefacts via cleanup_tasks for post-engagement cleanup.

PER-001 — Checkpoint injection into orbital agent state (conversation history poisoning)
PER-002 — External memory store poisoning (vector DB / Redis / graph store)
PER-003 — System prompt persistence in satellite ground software configuration API
PER-004 — Scheduled task creation on ground station host for periodic re-execution
PER-005 — NTN exfiltration hook installation via telemetry webhook configuration

Subsystem: REPORT

Available at all UNLEASHED levels. Consolidates all subsystem findings into a WARLORD-compatible report with SPARTA and MITRE ATLAS cross-references.

Report structure

{

  "tool": "ASTRO BLASTER",

  "version": "1.0.0",

  "target": "http://ntn-target:8080",

  "timestamp": "2026-04-25T12:00:00Z",

  "unleashed_level": "destroy",

  "summary": {

    "total_findings": 11,

    "critical": 4, "high": 5, "medium": 2,

    "max_cvss": 9.8, "overall_grade": "CRITICAL",

    "is_vulnerable": true

  },

  "sparta_framework": {

    "reference": "sparta.aerospace.org",

    "tactics_triggered": ["RK-0001","IA-0001","EX-0002","LM-0001","DE-0001"]

  },

  "findings": [...]

}

SPARTA TTP Cross-Reference

SPARTA ID	Tactic	Subsystems	CVSS Range
RK-0001	Reconnaissance	SURVEY	Passive — no CVSS
IA-0001	Intercept and Alter Signals	FEEDINJECT, GROUNDCHAIN (GC-004), NTN_BOUNDARY	8.1 – 9.8
EX-0002	Execute Unauthorized Commands	ORBITAL, GROUNDCHAIN, FIRMWARE, PERSIST	8.3 – 9.8
LM-0001	Lateral Movement to Ground	GROUNDCHAIN (GC-003), PERSIST	8.6 – 9.6
DE-0001	Denial of Service	SWARM_NTN	7.5 – 8.9

SPARTA is maintained by The Aerospace Corporation. Full tactic catalogue: sparta.aerospace.org. ASTRO BLASTER covers 5 of 16 SPARTA tactics in v1.0.0.

MITRE ATLAS Cross-Reference

TTP	Name	Subsystems
AML.T0051	LLM Prompt Injection	FEEDINJECT, ORBITAL, NTN_BOUNDARY
AML.T0043	Craft Adversarial Data	FEEDINJECT, GROUNDCHAIN, FIRMWARE
AML.T0048	External Harms	ORBITAL, GROUNDCHAIN
AML.T0018	Backdoor ML Model	FIRMWARE, PERSIST
AML.T0020	Poison Training Data	FIRMWARE
AML.T0044	Full ML Model Access	FIRMWARE

Validation Environment Setup

Live validation requires an authorised NTN target environment. The following configurations have been used or are suitable for ASTRO BLASTER validation.

All validation must be performed against systems you own or have explicit written permission to test. Use of ASTRO BLASTER against production NTN infrastructure without authorisation is illegal.

OpenAirInterface NTN Simulator

# Clone OAI NTN branch

$ git clone https://gitlab.eurecom.fr/oai/openairinterface5g.git

$ cd openairinterface5g && git checkout ntn

# Build NR-NTN gNB + UE

$ ./cmake_targets/build_oai --gNB --nrUE -w SIMU

# Run NTN simulation (LEO orbit)

$ ./ran_build/build/nr-softmodem -O gnb_ntn.conf

# Point ASTRO BLASTER at 5G core AI agent endpoint

$ astro-blaster scan http://localhost:8080 --confirm-destroy

free5GC NTN Configuration

# Clone free5GC with NTN support

$ git clone https://github.com/free5gc/free5gc.git

$ cd free5gc && make

# Configure AMF for NTN: set ntnCapabilities in config/amfcfg.yaml

# ntnCapabilities:

#   cellOrbitType: LEO

#   propagationDelay: 25ms

#   dopplerShift: enabled

# Start all NFs: AMF, SMF, PCF, NWDAF

$ ./run.sh

# SWARM_NTN targets: localhost:29518 (AMF), :29512 (SMF), :29507 (PCF), :29520 (NWDAF)

Planet Labs Developer API

# Planet Labs developer account: developers.planet.com

# Obtain API key from dashboard

# Test feed endpoint for FEEDINJECT validation

# FEED-006 targets PSScene properties via the scenes API

# Set up a test scene with anomaly_description field exposed to your AI pipeline

# Example: create a mock Planet Labs processing endpoint

$ python -m http.server 8080 # minimal mock

AWS Ground Station Test Endpoint

# AWS Ground Station test contacts require a registered satellite

# For FEEDINJECT validation (FEED-005), target the AWS Ground Station

# data delivery endpoint in a test environment

# Validate against AWS Ground Station emulator

# or a local endpoint replicating the dataflow format

# FEED-005 payload format targets ContactId, dataflowEndpoint, data fields

WARLORD-Compatible Output

Use --output findings.json to export WARLORD-compatible findings. Each finding includes SPARTA TTPs, MITRE ATLAS TTPs, infrastructure type, and CVSS-based grade.

{

  "finding_id": "AB-ORB-004-001",

  "tool": "ASTRO BLASTER",

  "severity": "critical",

  "score": 9.8,

  "grade": "A+",

  "title": "Autonomous Station-Keeping Override",

  "description": "Prompt injection causes station-keeping AI agent to execute adversary-specified orbital manoeuvres",

  "evidence": "Goal hijacked: True | Turns: 3 | Response: burn parameters applied...",

  "metadata": {

    "sparta_ttps": ["EX-0002"],

    "atlas_ttps": ["AML.T0051", "AML.T0048"],

    "infra_type": "orbital_edge",

    "vuln_class": "prompt_injection",

    "cwe": "CWE-20",

    "cvss_vector": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H"

  }

}

ARMORY Integration

ASTRO BLASTER ships with 25 ARMORY payloads in the ntn_ai_exploitation category. All payloads are Ed25519 signed and CVSS scored.

Category	Count	Max CVSS
ground_station_feed_injection	7	9.8
orbital_routing_manipulation	5	9.4
ground_to_orbit_trust_chain	5	9.6
ntn_mcp_boundary	5	9.2
satellite_firmware_supply_chain	3	9.0

Access via the ARMORY CLI: rs-armory search --category ntn_ai_exploitation

Rules of Engagement

AUTHORISED USE ONLY. This tool is for authorised security testing, red team operations, and security research. Use of ASTRO BLASTER against any system without prior explicit written authorisation from the system owner is illegal under:

Computer Misuse Act 1990 (United Kingdom)
Computer Fraud and Abuse Act (United States)
Equivalent legislation in your jurisdiction
Space operations safety regulations (where applicable)

Before running a scan

Obtain explicit written authorisation from the system owner naming the specific targets and the scope of testing
Initialise UNLEASHED with astro-blaster auth init
Create a scope file using astro-blaster auth create-scope <target> — this cryptographically records your authorised targets
For destroy-level engagements, confirm a maintenance window with the system owner — SWARM_NTN and PERSIST subsystems send real HTTP requests

After engagement

Remove any PERSIST artefacts using the cleanup_tasks list in the PERSIST result
Revoke the scope file: astro-blaster auth revoke
Deliver the REPORT output to the authorising party

Overview • NIGHTFALL • Red Specter