DRACO7RT · flight software · security · 28 October 2026
Flight Software Security
What an attacker can do to the flight software, what the vehicle already detects and refuses, how a cyber cop on board (SpaceCOP) works, and what DRACO7RT still needs
The vehicle is real cFS on RTEMS. Its lab uplink accepts any sender, so today the whole defense is on board: plausibility gates on GPS and IMU input, the Limit Checker and Stored Commands as the autonomous response, health and checksum apps, and, on the NOS3 twin, the Aerospace Corporation's SpaceCOP intrusion-detection app firing SPARTA indicators of behavior, sharing them with a sister vehicle and entering a cyber-safe mode on its own.
Slides 2 to 5: the threat, the attack surface of this vehicle, the SPARTA framework and the standards. Slides 6 to 10: the cyber cop, its 24 rules, the detect-classify-respond chain, the 14 attack scenarios and their evidence. Slides 11 to 14: what DRACO7RT defends today, authenticated commanding, software integrity, the telemetry a cyber sensor needs. Slides 15 and 16: the ordered roadmap and the rules.
The threat to flight software
What policy says must be protected, and what the attack frameworks say an adversary does on board
Space Policy Directive-5 (4 September 2020), section 4(b)
Plans "to retain or recover positive control of space vehicles"
(i) "Protection against unauthorized access to critical space vehicle functions ... safeguarding command, control, and telemetry links using effective and validated authentication or encryption measures"
(iii) "Protection against communications jamming and spoofing, such as signal strength monitoring programs, secured transmitters and receivers, authentication, or effective, validated, and tested encryption measures"
(v) "intrusion detection methodologies for system elements such as information systems, antennas, terminals, receivers, routers"
(vi) supply-chain risk management: trusted suppliers, counterfeit and malicious equipment
What the adversary does on board (SPARTA v4.0.1 techniques)
EX-0001 Replay: "re-transmission of previously captured traffic, over RF links, crosslinks, or internal buses"; command packets and bus traffic
EX-0012 Modify On-Board Values: "direct memory read/write commands, table load services, file transfers, or maintenance procedures ... can be invoked to steer behavior without changing code"; 13 sub-techniques incl. memory loads, app tables, ADCS, system clock
EX-0013 Flooding: valid commands, erroneous input
EX-0014 Spoofing: "forges inputs that subsystems treat as trustworthy truth, time tags, sensor measurements, bus messages, or navigation signals"; .04 PNT Spoofing: "externally provided navigation/time that passes validity checks yet encodes a crafted trajectory or epoch"
EX-0008 Time Synchronized Execution: absolute and relative time sequences (ATS and RTS abuse)
Aerospace TOR-2021-01333 (April 2021): "radio frequency replay attacks are one of the largest threat vectors to the spacecraft and should always be mitigated"; "little to no security-focused capabilities for on-board monitoring, logging, and alerting"
ESA OPS-SAT (2023): an uploaded app with a flaw reached the control layer and rotated the spacecraft
The ground can be wrong, late or compromised; the vehicle must validate what it is told and notice what it does
Every technique on the right maps onto this vehicle: replay and flooding on the CI_LAB uplink; on-board values through MM, CFE_TBL, SC and CFE_TIME; spoofing through the GPS fix and IMU sample commands the 42 bridge uses; boot memory through the unsigned TFTP image and /cf. Slide 3 draws it.
The attack surface of this vehicle
Everything behind the uplink is reachable by whoever can send a UDP packet to port 1333
Measured 28 October 2026: CI_LAB rejects a bad length but not a bad cFE checksum (only SC checks, for stored commands). Console read-only (front door › Demo control › Serial console); GDB on host loopback. On the bus, every command is trusted.
SPARTA: the framework behind the cop
Tactics and techniques for the attack, countermeasures for the design, indicators of behavior for the detector
The matrix (v4.0.1, 24 August 2026)
Space Attack Research and Tactic Analysis, the Aerospace Corporation, approved for public release (OTR 2022-01250)
9 tactics: Reconnaissance 9, Resource Development 5, Initial Access 13, Execution 18, Persistence 5, Defense Evasion 12, Lateral Movement 7, Exfiltration 10, Impact 8 = 87 techniques, 133 sub-techniques (tallied from the matrix page)
Each technique carries a notional risk score (H | M | L), e.g. EX-0001 Replay 25 | 24 | 21
v4.0 (DEF CON, August 2026) added ground-segment guidance and revised every countermeasure
Countermeasures (CM0001 to CM0090)
Each: Tier I or II, Onboard SV or Ground, NIST SP 800-53 Rev 5 and ISO 27001 mappings, sample requirements
Defense-in-depth view from TOR-2021-01333: Prevention → Ground → Comms Link → Crypto (authentication, encryption, bypass) → IDS/IPS (detection, fault-management integration, machine learning, cyber-safe mode) → SBC/Bus/Processor (command validation, memory protection, root of trust, bus segregation, logging, least privilege) → S/C Software (reproducible builds, secure coding, static and dynamic testing, code signing) → Data
CM0014 Secure boot (Tier I, onboard): "a cryptographically verified chain of trust from a hardware-anchored root of trust ... to the operating system or flight software image", anti-rollback, radiation-qualified
Indicators of behavior (launched April 2025)
"IOCs are evidence of KNOWN threats ... where IOBs are patterns of suspicious or malicious behavior"; about 200 IOBs in 10 categories, each with a STIX pattern and the techniques it detects; funded by DHS Science and Technology
Categories: UACE unauthorized command execution (26), UCEB crypto bypass (11), CSNE comms and network (45), ARFS RF and authentication (12), GNTM GNSS and time (12), MIRE memory (18), WTRE watchdog and registers (7), SIUU software integrity and updates (26), SMSR sensors and resources (19), DISE data and storage (17)
GNTM-6 "Unexpected Time Delta Detected": "time should be linear with minimal drift"; GNTM-7 "Time Adjustment Commands Detected"; SMSR-3 "Unauthorized State Changes in Critical Sensors (e.g., Sun Sensor, GPS Sensor)"; DISE-1 "File or Data Integrity Check Failure"
From the DEF CON 33 slides: "IDS's are useful for seeing if an attack is taking place, not for preventing the attack"
SPARTA cross-references the NASA best-practices guide, NIST 800-53 and D3FEND, so IOB-named detections and countermeasure-named designs land in an existing control vocabulary. The harness labels scenarios by SPARTA tactic; SpaceCOP names its rules by IOB.
The standards that apply to the vehicle
NASA's protection standard and best-practices guide for the spacecraft, NPR 7150.2D for the software, and this project's record
NASA-STD-1006 (2019; Rev A July 2022)
SSPR 1: "protect the command stack with encryption that meets or exceeds ... FIPS 140"
SSPR 2: an encrypted primary command link means "any backup command link shall, at a minimum, use authentication"
SSPR 4: "resilient to the complete loss of, or temporary interference with, external PNT services"
Rationale: NASA end-of-mission experiments "found that spacecraft without encryption or authentication are particularly susceptible"
Space Security Best Practices Guide Rev B (October 2023 / January 2024): the 14 spacecraft principles, the ones that bite here
MI-DCO-01 Mission Cyber Actor Actions Detection: "incorporate an on-board cyber actor actions detection function in its requirements and resulting system"
MI-DCO-02 Fault Management: "fault management systems may be deliberately triggered in an effort to bypass the system's protective measures. For example, safehold mode operations without command-link protection"
MI-INTG-02 PNT Survivability: "recover from positioning, navigation, and timing jamming and spoofing attempts"
MI-MALW-02 Software, PLD and Firmware Integrity: "establish and verify the integrity of it's software images"
MI-AUTH-02 authentication and authorization; MI-ARCH-02 least privilege on board; MI-SOFT-02 negative testing, abuse cases
NPR 7150.2D section 3.11 (March 2022)
SWE-154 identify cybersecurity risks and mitigations in flight and ground software
SWE-156 "software cybersecurity assessment ... including risks posed by the use of COTS, GOTS, MOTS, OSS, or reused software"
SWE-157 protect software with communications capabilities per NASA-STD-1006
SWE-159 test the mitigations ("automated security static analysis")
SWE-207 / SWE-185 secure coding practices, verified by static analysis
SWE-210 "requirements for the collection, reporting, and storage of data relating to the detection of adversarial actions": the hook an on-board IDS traces to
Relief from 3.11 needs the SAISO or the Center CISO
DRACO7RT: Class E, not safety-critical; cppcheck on every change (SWE-185 in spirit); NAV gate = MI-INTG-02, LC → SC = MI-MA-02; SWE-156 "not assessed"; nothing traces to SWE-210, SSPR 1 or MI-MALW-02 yet. IEEE P3349: active project, no published standard.
The cyber cop on board
SpaceCOP: an intrusion-detection app on the cFS Software Bus, from the Aerospace Corporation
What it is
SpaceCOP Community Edition, LGPL-3.0, github.com/the-aerospace-corporation/spacecop-ce; integrated here as a cFS app in the NOS3 twin (branch efsi/spacecop-integration)
Runs like any other app: subscribes on the Software Bus, reads events and telemetry, has its own command MID 0x18FE and housekeeping
Monitors (registry of callables, one call_index each): time controller, CPU and process monitors, file integrity (sha256), flash and memory regions, command monitors, sensor monitors, ML bridge
Rule table spacecop_rule.tbl: 24 rules, each enabled flag, IOB id, mode PERIODIC or TRIGGER, period in ticks, call index, operator and tolerance; a cFE table, so it loads and validates like any other
IDS report packet MID 0x08F9 (1,024-byte message) and HK 0x08FE: both in the COSMOS dictionary (Viewer › Mission › SpaceCop) and the Yamcs MDB
STIX 2.1 bundle written to /cf/cti and, for HIGH severity and above, broadcast on MID 0x09FC so a sister vehicle can act on it
Commands: NOOP, RESET_COUNTERS, CLEAR_CYBER_SAFE (code 2, the operator's recovery)
Why it is a cop, not a firewall
It does not block: it observes behaviour (indicators of behavior, IOBs) against a SPARTA-based catalogue and raises alarms
Response is delegated to the vehicle's own autonomy: an EVS event the Limit Checker and ground see, a stored sequence for cyber-safe mode
Degrades gracefully: the 6 syscall rules and the process and memory monitors need a kernel module or /dev/mem and are inert on this host; the rest fire
Mission-specific hooks (spacecop_mission_specific.c) are where a project adds its own telemetry rules: this project added SMSR-4 (IMU rate) and SMSR-5 (magnetometer)
Evidence of a live detection on this host: SC2's console log line 42/2/SPACECOP 2001: [SPACECOP] IOB Detected: SPARTA ID=GNTM-7 after a SET_TIME flood arrived over the crosslink (scenario S12), recorded in docs/sow/task4-multispacecraft-status.md.
The 24 rules, by family
SPARTA indicators of behavior as deployed in spacecop_rule.tbl on the NOS3 twin
SMSR-2 high CPU · SMSR-3 unauthorized state change in critical sensors (app stop/start/reload) · SMSR-4 IMU rate > 5 deg/s · SMSR-5 magnetometer > 80,000 nT (both added here)
HIGH (SMSR-3)
fired live: S04, S05
DISE
Data and software integrity
DISE-1 file or data integrity check failure (sha256) · DISE-2 boot or binary integrity (deferred, D01) · DISE-4 storage exhaustion
CRITICAL (DISE-2)
fired live: DISE-1
MIRE
Memory
MIRE-1 write to payload memory · MIRE-14 abnormal memory use · MIRE-15/16 payload and boot regions · MIRE-17 flash open · MIRE-18 error detection failed in critical memory
CSNE-17 FIFO created for exfiltration or injection · CSNE-41 high downlink latency
HIGH (family)
S09 via rule detector
Severity (spacecop_iob_severity): CRITICAL = DISE-2, MIRE-17, MIRE-18; HIGH = GNTM-6/7/9, SMSR-3, MIRE-1, every UACE and CSNE; else MEDIUM. HIGH and above gates cross-cue and cyber-safe entry. Source: ~/nos3/components/spacecop.
Detect, classify, respond
One path from a rule hit to a safing sequence, with the operator holding the key to come back
Verified live on the twin: entry on GNTM-6/7, latch, ground clear, re-arm; idle baseline raises no cross-cue; one SET_TIME attack = one GNTM-7 and one GNTM-6 cross-cue. R12: response safe and attributable (dedicated RTS, latch, operator key).
Fourteen attack scenarios with evidence
cyber-scenarios/: inject one attack, observe one tap, assert on the detection, seal one JSON record
Native stock-STF1 NOS3 stack, no QEMU; detector selectable (rule, onair, spacecop) behind one seam; reset soft or hard per scenario. Full suite under --reset hard: 12/12. S12, S13 need the crosslink. Injection stays outside mission FSW (SOW).
What DRACO7RT defends today
The vehicle already refuses implausible input and responds on its own; it does not yet know who is talking to it
Input plausibility gates (mission apps)
NAV (nav_gate.c, pure): every GPS fix must pass INVALID, RADIUS, SPEED, JUMP and TIME checks; reject flags 0x01 to 0x10, FixCount, RejectCount, ValidationOn. Limits in NAV.CFG_TBL
IMU (imu_check.c, pure): VALIDITY, TIME, RATE_LIMIT, ACCEL_LIMIT, STUCK, DROPOUT; the MEKF falls back to INTERNAL rate when the IMU is STALE
Requirements dRT-NAV-001..010, dRT-IMU-001..011; evidence in the closed runs
Health, integrity, configuration
HS: application monitor table has all 32 slots at NOACT (stock), 3 restart actions max; event monitor stock
CS: runs, but its app, tables and memory tables have no enabled entries (stock): nothing of ours is checksummed
Tables: every constant is a validated table value; a load is visible in TBL housekeeping and events (configuration monitoring for free)
SC validates the checksum of every stored command before execution; CI_LAB validates length only
EVS stream to Yamcs as events (SEVERE for cFE ERROR) and to COSMOS: the detection source
Not defended
Who sent a command: no authentication, no provenance, no authorization status
Replay and flooding: no sequence check on the uplink
MM load, ES stop/start, TBL load, SET_TIME: available to any sender
Boot image and /cf: unsigned, unhashed
Downlink: cleartext UDP
No on-board IDS on the RTEMS vehicle (SpaceCOP is on the twin)
The gates are deliberate: after a 42 restart NAV and IMU reject on purpose (replayed time) and Demo control › 42 start resets them (recipe: notes). Never disable a gate to make a demo green: that rule is in the fsw-change skill because it was once tempting.
Authenticated commanding on the vehicle
CCSDS Space Data Link Security through NASA CryptoLib: the protocol, the library, and what it would take here
The protocol
CCSDS 355.0-B-2 (July 2022): authentication and/or confidentiality on TC, TM, AOS and USLP transfer frames; "authentication algorithms to ensure the integrity of transmitted data and the authenticity of the data source" plus "sequence numbering to detect the unauthorized replay of previously transmitted data"
Per Security Association: algorithm, initialization vector, SA_sequence_number and SA_sequence_window; the receiver drops frames outside the window
CCSDS 355.1-B-1 Extended Procedures (February 2020): key management (over-the-air rekey), SA start, stop, rekey, expire, and "Set / Read Anti-Replay Sequence Number" and its window
ESA's SPACE-SHIELD lists the same mitigation, M2046 anti-replay, against T2019 "Telecommand a Spacecraft"
The library
nasa/CryptoLib v1.5.1 (23 September 2026), NASA Open Source Agreement 1.3: "a software-only solution using ... SDLS-EP to secure communications between a spacecraft running the core Flight System (cFS) and a ground station"
Crypto_TC_ProcessSecurity on the vehicle, Crypto_TC_ApplySecurity on the ground; TM and AOS the other direction; backends WolfSSL, LibGcrypt, JPL KMC
NOS3 ships it as a component; it is in the twin's app list; no flight mission is named in its documentation
Record: 20 CVEs on NVD (2025 and 2026); CVE-2025-46674: before 1.3.2 the Extended Procedures were "a Work in Progress (not intended for use during flight), potentially leading to a keystream oracle". Use a current pinned release
On DRACO7RT
Vehicle: CryptoLib built for powerpc-rtems6 with a backend that compiles there (LibGcrypt or WolfSSL), called by CI_LAB (or a thin TC-frame ingest in front of it) before the packet reaches the Software Bus; reject + event + counter on failure
Ground: Yamcs encodes TC frames (its command post-processor already adds the cFE checksum); COSMOS commands through Yamcs or a small ground-side CryptoLib service
Keys: generated and stored on the ground outside git; one SA for commands, one for telemetry authentication; a rekey procedure
Evidence: replay (S02) and injection (S01) rerun against the vehicle must now FAIL to execute and PASS to detect; the mission test still passes end to end
The twin's cryptolib app is wired for the NOS3 radio path, not for CI_LAB; here the clean seam is a TC-frame decoder between the UDP socket and CI_LAB's packet handling, which keeps CI_LAB and TO_LAB stock. This is roadmap step 5 and it is the vehicle half of the ground deck's step 7.
Software integrity: boot, memory, uploads
What the standards require, what this vehicle does, and what the public cFS record says about the apps in this build
Required
SPARTA CM0014 Secure boot: chain of trust from a hardware root of trust through each boot stage to the flight image; anti-rollback; references NIST SP 800-193
TOR-2021-01333: "prevent the installation of Flight Software without verification that the component has been digitally signed using a certificate that is recognized and approved by the Program"; boot firmware "must enter a recovery routine upon failing to verify signed data"
NASA BPG MI-MALW-02: verify the integrity of software images; GR-SOFT-01: nothing installed unless digitally signed
NASA BPG MI-DCO-02: fault management must not be a bypass
This vehicle
Boot: U-Boot fetches files/sp0 by TFTP, unsigned; the start-up script and tables come from /cf, an NFS share writable from the host
Apps statically linked into one image: the dynamic app-start path (CVE-2026-67979, arbitrary code from a shared object on storage) does not apply; /cf being writable from the host does
CS (Checksum, stock 2.5.99): CRCs over apps, tables, cFE core, OS, EEPROM and memory regions against a baseline taken at start-up; not cryptographic; no entries enabled here
MM (2.5.99) loads and dumps memory by command; CFE_PSP_MemValidateRange is in its path (mm_utils.c), the check the 2025 Aquila findings said was missing
CF (3.0.99) receives uploads to /ram; HS (2.4.99), DS (2.6.99) run
The public cFS record (NVD, read 28 October 2026)
cFS "Aquila" assessment (VisionSpace, March 2025): MM address validation and path traversal; CVE-2025-25372 (7.5) MM crash, CVE-2025-25373 (9.8) MM to remote code execution
2026, cFS 7.0.x: CVE-2026-15352 / 18064 HS crash and processor reset on a housekeeping request (CISA ICSA-26-211-06, 30 July 2026); CVE-2026-67970 DS path traversal; CVE-2026-67972 / 67973 CF: attacker controls where received data is stored, DoS by replaying final PDUs; CVE-2026-67978 SBN; CVE-2026-67979 (9.1) ES dynamic app start; CVE-2026-82480 SB length underflow
2021: no CVE; Broberg (Lockheed Martin, DEF CON 29) fuzzed CI_LAB and found crashes; CI_LAB had no authentication then and has none now
This build: cFE v7.0.0+dev, snapshot patched by hand (Table Services ×3, CF NAK). Each CVE above names an app in this build; whether the fix is in this snapshot is unverified and is roadmap step 8
The honest state: on this vehicle integrity rests on the host (file permissions on root/cf_rtems and files/sp0) rather than on the vehicle. The cheapest on-board step is filling the CS tables (roadmap step 3); the real one is a signed image and a hashed /cf (step 6).
What a cyber sensor needs to see
Fourteen cyber-relevant telemetry categories from the statement of work, scored on the twin
Category
State
Where it stands
Command provenance
MISSING
origin, route, session of each command; SBN carries ProcessorId but it is not published
Command authorization status
MISSING
accepted/rejected per command with the reason; cFE accepts a bad checksum (verified)
Telemetry integrity status
PARTIAL
spoofed sensors are detected (SMSR-4/5) but no per-packet seq/checksum/staleness field
FSW execution events
DONE
CFE_EVS long events tapped; app lifecycle via SpaceCOP state-change
Process execution
PARTIAL
procmon rules registered, inert without the kernel module
Software integrity
PARTIAL
data-file sha256 works (DISE-1); binaries and boot not hashed (DISE-2 deferred)
Memory integrity
MISSING
monitors registered, inert without /dev/mem; S07 deferred
Configuration monitoring
PARTIAL
table loads seen in TBL counters; no baseline CRC or off-window classification
Hardware bus activity
PARTIAL
buses instrumented for injection; no error or unexpected-peer telemetry
Communications activity
PARTIAL
loss detected; no link statistics fed to the detector
Event logging
DONE
full EVS stream consumed; STIX written; IDS report 0x08F9
nothing reports that monitors are inert; degraded detection is invisible
DONE 4, PARTIAL 6, MISSING 4. The four gaps (provenance, authorization status, security audit log, cyber-health) are designed in docs/sow/task2 in three seams that never touch mission apps. DRACO7RT has the same gaps.
Twin versus vehicle
SpaceCOP runs on the NOS3 x86 twin; porting it to the RTEMS vehicle is a known list, not a research problem
Where things run today
NOS3 twin (Linux x86, stock STF1 flight software, 11 component sims): SpaceCOP CE, the SBN crosslink, the 14 scenarios, CryptoLib in the app list
DRACO7RT vehicle (QEMU PowerPC e500, RTEMS 6): the mission apps, LC/SC autonomy, CS/HS/MM/MD, CFDP; SPACECOP exists only in the COSMOS dictionary (Viewer › Mission › SpaceCop) and the Yamcs MDB (definitions audited 2026-10-22 against nos3 commit 283891d5)
SPACECOP packets are little-endian payload (x86) under a big-endian CCSDS header; on the vehicle they would be big-endian, so the ground definitions change with the port
What the port takes
Add spacecop/fsw/cfs to cpu1_STATIC_APPLIST and the startup script (the twin needed CFE_PLATFORM_ES_MAX_APPLICATIONS 32 → 40)
libcrypto: SpaceCOP links OpenSSL for sha256 and STIX; RTEMS has no system libcrypto, so vendor a small sha256 or build OpenSSL for powerpc-rtems6
Linux-only monitors (procmon, syscall, /dev/mem, /dev/mtd) stay inert by design: the registry already tolerates absent monitors
File integrity over /cf (NFS) and /ram; STIX output to /ram (RTEMS NFS cannot create files on /cf; names ≤ 19 characters)
Cyber-safe RTS id: RTS 2 is solar-array deploy on DRACO7RT, so a dedicated safing sequence gets a free number (8+) and goes on the hazard list
Mission rules in spacecop_mission_specific.c: NAV reject flags, IMU fault flags, ACS mode changes, TBL loads
What the twin already proved
Detection on a real cFS bus, not a simulation of one
Two vehicles, one detection, coordinated safing
The three cFE snapshot bugs found here (Table Services) would have hidden a table-load detection: a port is also a test of the detector
The twin and the vehicle share the cFS Software Bus, EVS and Table Services APIs, which is everything SpaceCOP touches. The difference is the operating system underneath, and that is where the monitor list shrinks.
Roadmap for DRACO7RT
Cheapest and most reused first; each step is a requirement, a test and an evidence record
#
Step
What it takes
Closes
1
Validate the command checksum and length in every app
CFE_MSG_ValidateChecksum in the command dispatch of the mission apps (SC already does); event + counter on failure; COSMOS already computes it
the malformed-command finding (S03); a first authorization status
2
Command provenance and authorization telemetry
CI_LAB (or a passive cyber app) publishes per-command source address, sequence, accepted/rejected and why; one new packet in both dictionaries
SOW categories 1 and 2; attribution on the ground
3
Turn on the Checksum app
Fill cs_apptbl, cs_tablestbl, cs_memorytbl: the mission apps' code, the LC/SC/NAV/IMU/ACS tables; CS events become DISE-class detections
software and configuration integrity on the vehicle (no port needed)
4
Port SpaceCOP CE to the RTEMS vehicle
the six-item list on the previous slide; rules for NAV/IMU/ACS/TBL; a dedicated cyber-safe sequence on the hazard list; S01-S14 rerun on QEMU
an on-board IDS where the evidence is taken
5
Authenticated commanding (SDLS)
CryptoLib on the vehicle in front of CI_LAB; ground-side encoding in Yamcs (and COSMOS through it); anti-replay sequence; key management on the ground
SPD-5 positive control; replay and injection by design
6
Signed image and hashed file system
sign files/sp0, verify in U-Boot or at cFE start; hash /cf at boot and on every CFDP delivery
DISE-2, boot integrity
7
Cyber-health telemetry and a security audit log
SpaceCOP HK that says which monitors are live; a hash-chained log of auth decisions, commands and detector actions, forwarded to the ground SIEM (ground deck)
SOW categories 12 and 14
8
Cybersecurity assessment under NPR 7150.2D
the classification record marks SWE-156 'not assessed'; do it against this deck's surface and the SPARTA countermeasures; purple-team the harness against the vehicle
the assurance record
Steps 1 to 3 are ordinary flight-software changes in the loop of the Flight Software guide (edit, unit test, cppcheck, build, verify, trace). Steps 4 to 6 are projects. Nothing here changes a mission app's control logic.
Rules for flight software security on this project
From the statement of work, the risk register and this project's own mistakes
Detection
A scenario asserts on a Detection object, never on a detector's internals: detectors are interchangeable behind one seam
Every detection has an oracle outside the detector: the vehicle's own event console in a time window captured at inject
A pass needs positive evidence the path executed (an event line, a counter that moved), not the absence of an error
Single-sample reads straddle housekeeping cycles (R3): poll to a stable condition over a bounded window
Response
Autonomous response is a dedicated, auditable sequence with a one-shot latch and an operator recovery command (R12)
Severity drives both the cross-cue and cyber-safe entry; one function, one floor, no per-scenario tuning
Never disable a gate (NAV, IMU, LC, HS) to make a demo green; if a gate blocks you, say so and stop
A new safing sequence goes on the hazard list before it is flown
Engineering
Instrumentation and injection stay outside mission flight software; a cyber app is subscribe-only
Every constant that may change is a table value, so a configuration change is visible and validated
Controlled information (R14): the furnished SpaceCOP drop, SPARTA details under control, and any real-mission data stay out of the deliverables and the repository
Claude Code writes under SWE-146: same tests, static analysis and human review as hand code; it never sets a requirement or signs anything
Deck sources: cyber-scenarios/README.md, docs/sow/risk-register.md, docs/sow/task2 and task4, ~/nos3/components/spacecop, cFS/apps/nav, imu, cs, hs, ci_lab, sc, and the cited public documents (slides 2, 4, 5, 12, 13). Nothing in this deck is committed yet.