Burak Oktenli

Research Mission

The Problem. Autonomous systems in defense and civilian transportation make decisions faster than humans can intervene, yet no standardized architecture enforces real-time authority governance. DoD Directive 3000.09 requires human control over autonomous weapons but lacks a technical enforcement mechanism. Nearly 40,000 Americans die annually in traffic crashes; NHTSA has documented 1,429 AV incidents (2021-2025); and the SELF DRIVE Act of 2026 (H.R. 7390) now mandates cybersecurity plans for "false vehicle control commands", but no validated reference architecture exists to implement these requirements.

The Gap. Policy frameworks (NIST AI RMF) establish principles but do not enforce them in hardware. Planning-layer safety systems (Mobileye RSS, NVIDIA SFF, SOTIF) constrain behavior but do not govern whether commands reach actuators. Software-only monitors can be bypassed. What is missing: a hardware-enforced architectural layer that continuously evaluates sensor trust, computes graded authority, enforces deliberation windows, and cuts actuator authority through a fail-safe when trust degrades, deterministically, in real-time, without firmware involvement.

My Contribution. Seven governance architectures managing the full authority lifecycle: trust evaluation, authority computation, command gating, consensus, deliberation, deception detection, and recovery:

• SATA: Sensor trust attestation using weighted Dempster-Shafer fusion with cross-sensor validation (Patent: U.S. Provisional 64/002,453)
• HMAA: Four-level authority computation (A3-A0) with asymmetric hysteresis: immediate downgrade, 5-15s delayed upgrade (Patent: 63/999,105)
• CARA: Deterministic GREP-phase recovery: Govern → Restrict → Execute → Persist, with mutual exclusion (Patent: 64/000,170)
• MAIVA: Byzantine fault-tolerant multi-agent consensus with CUSUM-augmented anomaly detection
• FLAME: Mandatory deliberation windows preventing safety-critical actions without sustained authority (Patent: 64/005,607)
• ADARA: Adversarial deception detection computing P(adversarial) from cross-sensor consistency and temporal anomalies
• ERAM: Escalation risk quantification for AI-enabled command-and-control decision compression

These are documented in four U.S. provisional patent submissions, seventeen DOI-registered Zenodo records (including research papers with full simulation data for the rover testbed, UAV platform, BLADE-EDGE defense platform, BLADE-AV automotive platform, BLADE-MARITIME maritime platform, and BLADE-INFRA critical infrastructure platform), and fifteen SSRN papers on AI governance and national security policy. Total: 36 published works with public DOIs and permalinks. The BLADE-SPACE orbital governance node (TRL 2-3 Preliminary Design, NASA SBIR EXPAND.3.S26B aligned) ships with a 15-document engineering package and is published on Zenodo (DOI 10.5281/zenodo.20183269).

Hardware Implementation. Twelve physical platforms demonstrating governance at increasing operational capability:

• Rover Testbed, 37 components, ~$484, SATA-HMAA-CARA pipeline, 350 simulation runs across 7 fault scenarios, TLA+ verified (48,751 states, 8 safety properties). DOI: 10.5281/zenodo.19143190.
• UAV Platform, 52 components, ~$4,200, MAVLink/HIL bridge for Cube Orange+ integration, 250 simulation runs across 5 adversarial scenarios, Monte Carlo campaign. DOI: 10.5281/zenodo.19128769.
• BLADE-EDGE, 72 components, ~$139K, defense-grade directed-energy weapon governance. Dual Jetson AGX Orin + dual Zynq UltraScale+ FPGA. 9-module pipeline, MIL-STD-810G, hardwired safety interlock relay. DOI: 10.5281/zenodo.19177472.
• BLADE-AV, 62 components, ~$16K, autonomous vehicle drive-by-wire governance. 9-module pipeline, three-leg redundant KILOVAC LEV200 fail-safe relay. ISO 26262 ASIL-D, SAE J3016 L4. 1,200 runs, no observed unsafe authority transitions. DOI: 10.5281/zenodo.19232130.
• BLADE-MARITIME, 84 components, ~$43K, maritime surveillance governance. 9-module pipeline with hydroacoustic sonar, MAD, AIS spoofing detection, and sea-state authority damping. IP68 / MIL-STD-810G. DOI: 10.5281/zenodo.19246785.
• BLADE-INFRA, 92 components, ~$12K, critical infrastructure governance. 9-module pipeline with ICS/SCADA integration (IEC 61850, Modbus, PROFINET). SIL 3 / NERC CIP / FIPS 140-2. DOI: 10.5281/zenodo.19277887.
• BLADE-SPACE, 91 components, ~$505K, orbital governance node (TRL 2-3 Preliminary Design Phase). 9-stage pipeline on rad-tolerant compute (Microchip RTG4 FPGA + Aitech S-A1760 Venus SBC, hot-redundant); LEO 400-1200 km, 30 krad TID, 5-year design life; ECDSA P-256 audit chain; three-fault-tolerant safety interlock. 15-document engineering package. Aligned with NASA SBIR EXPAND.3.S26B. Zenodo DOI 10.5281/zenodo.20183269.
• BLADE-CUAS, counter-UAS authority governance node, ~$43.5K reference BOM (TRL 2-3 hardware / 3-4 simulation). 9-stage AUTHREX pipeline; four-tier HMAA (T3/T2/T1/T0) with federal-SLTT handoff; Dempster-Shafer consensus across radar, RF, EO/IR, Remote ID, LIDAR; ECDSA P-256 court-admissible evidence chain (Fed. R. Evid. 901/902/803(6)). Sixth BLADE platform, ~75% reuse from BLADE-EDGE. Aligned with EO 14305 and the FY26 NDAA Safer Skies Act. Zenodo DOI 10.5281/zenodo.20299604.
• BLADE-AGENT-HSM, agentic-AI hardware root of trust, ~$199 reference BOM (TRL 2-3 silicon / 3-4 emulator). Hardware companion to the AUTHREX-AGENT software shim; non-exportable ECDSA P-256/P-384 keys in a CC EAL6+ secure element; authority-tier state in a TPM 2.0 PCR bank; HKDF per-tool tokens; multi-modal tamper cascade. Five-opcode 64-byte ABI; USB-A stick and M.2 Key-E module. Seventh BLADE platform, first hardware root of trust. Aligned with CISA/NSA/Five Eyes agentic-AI guidance and FY26 NDAA Sections 1513 and 6601. Zenodo DOI 10.5281/zenodo.20299821.
• BLADE-SWARM, attritable swarm authority governance, ~$1,333 per node reference BOM (TRL 3-4 simulator / spec, TRL 2 testbed). Byzantine-fault-tolerant two-phase consensus gated by SATA/HMAA/MAIVA across N=10/50/500 agents; tolerates f=(N-1)/3 compromised agents per quorum with a quorum-intersection bound; per-node ECDSA P-256 root of trust and hash-chained distributed audit ledger; TLA+ verified (5 safety, 3 liveness). Eighth BLADE platform; the multi-agent extension of the HMAA-UAV platform. Aligned with DoDD 3000.09, FY26 NDAA, and the NIST AI RMF. Zenodo DOI 10.5281/zenodo.20351198.
• BLADE-INFRA-OT, authority-governed IT/OT bridge for cross-boundary OT command adjudication, ~1U fanless reference design (TRL 2-3 hardware / 3-4 simulation). Fail-closed, bump-in-the-wire governance appliance at the IT/OT segmentation boundary; AUTHREX adjudicates each cross-boundary command to propagate, hold, or isolate across four OT authority regimes; Xilinx Kria K26 governance plane and x86 network plane; 48 BOM line items, 35 electrical, 42 mechanical; seed-deterministic SHA-256 tamper-evident audit ledger. Ninth BLADE platform; the operational-technology companion to BLADE-INFRA. Aligned with NIST SP 800-82, ISA/IEC 62443, and NERC CIP. Zenodo DOI 10.5281/zenodo.20342067.
• BLADE-FINANCE, authority governance for financial-sector AI decision systems under the U.S. Treasury Financial Services AI Risk Management Framework (TRL 3-4 simulation / TRL 2 hardware). Software-enforced authority-arbitration node; an eight-stage AUTHREX pipeline (VALIDATE through CARA) routes each transaction to autonomous clearance, supervised review, elevated confirmation, or manual hold; four-tier HMAA; a population-state coordination model across account, device, payee, and IP-cluster history; a retrospective stigmergic swarm-review module; SHA-256 canonical-form evidence chain; 36 components, 33 electrical, 32 mechanical, approximately $9,228 BOM. Tenth BLADE platform; first in the economic-security domain. Aligned with the U.S. Treasury FS AI RMF, NIST AI RMF, and EO 14179. Synthetic data only; not deployed in any financial institution. Zenodo DOI 10.5281/zenodo.20374692.

All ten BLADE platforms share the same governance pipeline architecture, demonstrating domain-agnostic portability across defense (DoDD 3000.09), automotive (ISO 26262), maritime (MIL-STD-810G), critical infrastructure (SIL 3 / NERC CIP), orbital (NASA EXPAND.3.S26B alignment), counter-UAS (EO 14305), and agentic AI (CISA/NSA/Five Eyes agentic-AI guidance). Combined across the four flight-relevant published platforms: 2,800+ simulation runs, no observed unsafe authority transitions. BLADE-SPACE is at TRL 2-3 (Preliminary Design); its V&V campaign is specified but not yet executed.

Software Implementation. A companion software authority-governance layer on the AUTHREX Systems venture site extends the same authority pipeline to additional settings, as six reference architectures (TRL 3-4, single-author research, not deployed): AUTHREX-AGENT (agentic AI), AUTHREX-ASSURE (pre-deployment assurance), AUTHREX-ICS-GATE (OT and critical infrastructure), AUTHREX-AGENT-CYBER (cyber-defense, governance only), AUTHREX-SPACECYBER (orbital autonomy), and AUTHREX-SANDBOX (test and evaluation). Several are software companions to the BLADE hardware platforms.

Simulation and Verification. Nineteen browser-based simulations with seeded PRNG for bit-exact reproducibility. G*Power justified sample sizes, Bonferroni correction, Shapiro-Wilk normality tests. All run client-side with zero dependencies, any reviewer can independently verify results.

Current Status. Independent Researcher based in Washington, DC, pursuing Georgetown University M.P.S. Applied Intelligence (STEM, in progress). B.Sc. Computer Science Engineering (USF, 3.45 GPA). MBA International Business (Lynn, 4.0 GPA). Governance architectures published on Zenodo under CC BY 4.0; ERAM published on SSRN. Next-phase: physical validation, TLA+/UPPAAL formal verification, ROS 2/Gazebo HIL testing, and ASIL-D certification pathway. This research program is informed by progressive professional experience across data governance, infrastructure security, and critical infrastructure systems, see Professional Background below, and sustained engagement with the defense, aerospace, and AI communities through industry conferences and professional societies.