Strategic Capitalization of Automotive Software and ‘Secure-by-Design’ Strategy Research

Introduction : The Paradigm Shift in the SDV Era and the Fear of Practical Implementation

In the traditional embedded development field and the overall automotive industry, the view of software as merely a hardware accessory or subordinate control logic has completely disappeared.

As the vehicle’s electrical/electronic (E/E) architecture evolves from dozens of distributed Electronic Control Units (ECUs) to centralized High-Performance Computing (HPC) and Zonal architectures, software has been elevated to the most critical strategic asset determining a vehicle’s overall value, performance, and survival.

A key focus of recent global tech conferences is the reliability of lightweight AI models running at the vehicle’s edge and the implementation of advanced cybersecurity architectures to protect them.

Security standards such as UNECE R155 (Cyber Security Management System, CSMS) and R156 (Software Update Management System, SUMS) are becoming absolute prerequisites for new vehicle type approvals globally, starting in mid-2026.

Consequently, the ‘Secure-by-Design’ approach has transitioned from a mere recommendation to an essential norm for market launch.

Engineers are now fiercely debating how to efficiently integrate next-generation security technologies like Zero Trust Architecture (ZTA) and Post-Quantum Cryptography (PQC) into highly resource-constrained embedded environments.

Pain Point 1: Direct Collision with Legacy Compliance

The automotive industry has developed conservative and strict functional safety and quality management standards, such as ISO 26262, IEC 62279, and A-SPICE.

A major issue arises because these legacy standards assume deterministic software forms, which directly conflict with the probabilistic, non-deterministic “black box” nature of edge-based lightweight AI models.

Traditional safety standards demand complete tracking of systematic faults through the V-model lifecycle and perfect verification at each stage.

However, lightweight AI models process data using deep neural networks (DNNs), making internal decision-making processes impossible to intuitively track or reduce to specific source code.

This creates structural limitations in transparency, unit verifiability, and test scenario robustness, making it impossible to perform root-cause analysis or prove Automotive Safety Integrity Level (ASIL) achievements.

To bridge this gap, the industry introduced ISO/PAS 8800, which connects AI’s unique risks to traditional functional safety and SOTIF (Safety of the Intended Functionality, ISO 21448).

It redefines the AI element lifecycle, defining unintended results caused by data bias or environmental changes as ‘functional insufficiencies’. Furthermore, A-SPICE 4.0 was introduced to establish a “Machine Learning Engineering (MLE)” process group.

This standardizes AI development by mandating specific processes for ML requirement analysis, architecture design, training, testing, and data management.

Pain Point 2: Absence of Orchestration Architecture for Agent Coordination

The advancement of HPC, autonomous driving, and smart cockpits has led to a Multi-Agent ecosystem where multiple domain-specific edge AI models operate simultaneously.

However, the industry lacks an orchestration architecture to prevent conflicts among these agents and efficiently distribute workloads within limited embedded resources.

To orchestrate multi-agent systems, developers must choose appropriate architectural patterns, such as sequential pipelines, concurrent execution, or autonomous collaborative models (A2A), depending on functional requirements and latency constraints.

A critical difference between IT and automotive environments is the coexistence of “Mixed-Criticality”.

ASIL-D safety-critical models (like steering) and Quality Management (QM) level models (like infotainment) must operate simultaneously on a single HPC.

To solve this, the industry is focusing on open standards like the SOAFEE initiative. Structural solutions include containerization and hypervisor isolation to logically partition hardware resources, lightweight unikernels to minimize overhead, and standardized Agent-to-Agent (A2A) communication using sandboxing and protocols like the Model Context Protocol (MCP) to prevent unauthorized data exchange.

Realization of Strategic Asset Capitalization and 2026 ‘Secure-by-Design’

Even with functional safety and orchestration resolved, the strategic value of software collapses if it is not protected from cyberattacks.

With UN R155 and R156 regulations enforcing systematic security management across the entire lifecycle, manufacturers must adopt a ‘Secure-by-Design’ approach.

The technical foundation of this approach is Zero Trust Architecture (ZTA).

As attack surfaces expand through V2X and OTA updates, traditional closed-network security is entirely obsolete.

ZTA requires continuous identity verification, micro-segmentation with strict access control firewalls, and Hardware Root of Trust using Hardware Security Modules (HSM) and Secure Boot.

To overcome the resulting latency, architectures must incorporate hardware acceleration and multi-level trust caching.

Furthermore, the looming threat of quantum computing requires migrating to Post-Quantum Cryptography (PQC) by 2026 to prevent “Store Now, Decrypt Later” (SNDL) attacks that threaten firmware assets and IP.

Because PQC algorithms are complex and require larger memory footprints, the industry is developing ultra-lightweight modules and utilizing hybrid encryption for crypto-agility to facilitate this transition without sacrificing real-time responsiveness.

IP Monetization and Value Chain Reversal in the Automotive Industry

Once fully secured and orchestrated, software transforms into a strategic asset capable of generating massive recurring revenue through Over-The-Air (OTA) updates, Software-as-a-Service (SaaS), and Feature-on-Demand (FoD) models.

This causes a reversal in the value chain, shifting Intellectual Property (IP) strategies from mechanical parts toward autonomous driving algorithms, edge AI networks, software orchestration, and security protocols.

Companies now employ a two-track strategy: strictly protecting high-value core assets while actively utilizing open-source software (OSS) ecosystems for lower-differentiation, highly versatile areas to reduce development costs.

Conclusion: Architectures Reflecting Legacy Standards and Human-Centered Engineering

The automotive industry must overcome the fears of legacy compliance conflicts through ISO/PAS 8800 and A-SPICE 4.0 integration, and resolve agent orchestration via SOAFEE-based mixed-criticality architectures.

Implementing ‘Secure-by-Design’ through ZTA and PQC will serve as a shield protecting software asset value.

Ultimately, this requires ‘Human-Centered Engineering’.

No matter how advanced AI generation and orchestration become, the final evaluation of functional safety in extreme edge cases and the ethical judgment required to balance safety and convenience remain the profound responsibility of highly trained human engineers.