RTCA Releases New Avionics Safety & Security Standards
RTCA has released new electronic versions of two key industry standards, DO-354 and DO-325. DO-325 is critical for airborne systems security, a growing concern with networked and AI-enabled avionics, while DO-354 continues to guide safety assurance for electronic systems. The releases signal an ongoing evolution in regulatory requirements for modern aircraft.
DO-325, officially the "Minimum Operational Performance Standards (MOPS) for Automatic Flight Guidance and Control Systems and Equipment," provides the baseline for certifying systems like autopilots and auto-throttles. Its counterpart, DO-354, focuses on "Safety and Performance Requirements for CDTI Assisted Visual Separation (CAVS)," which allows aircraft to safely reduce separation based on ADS-B data displayed in the cockpit. The drive to update these standards stems from the increasing interconnection of aircraft systems. As avionics become nodes in a larger network, they are exposed to cybersecurity threats that were not a primary concern when the original documents were published. This has led to the development of a complementary set of standards, including DO-326A, the "Airworthiness Security Process Specification," which provides a framework for identifying and mitigating cyber threats to maintain safety. This security framework runs parallel to the safety-oriented processes defined in the well-established DO-178C for software and DO-254 for hardware. While DO-178C ensures a system functions as intended without failure, the DO-326 series ensures it can withstand unauthorized electronic interaction. The FAA is increasingly requiring manufacturers to address these security aspects as a condition of certification. The introduction of artificial intelligence and machine learning (AI/ML) presents the next major challenge for avionics certification. Existing standards are not designed for non-deterministic systems where behavior is learned from data rather than explicitly programmed. The FAA acknowledges this gap and in August 2024 released its "Roadmap for Artificial Intelligence Safety Assurance," signaling an incremental approach to certifying AI, starting with lower-risk applications. This roadmap outlines the FAA's plan to develop policies for assuring the safety of AI in aircraft, with draft policies anticipated in late 2024 and official certification position papers expected in the first quarter of 2026. Key principles include treating AI as a tool, not a person, and differentiating between static "learned" AI and dynamically "learning" AI in flight. The challenge for the industry and standards bodies like RTCA is to develop methods for the verification and validation of AI/ML systems that can provide the same level of confidence as for traditional systems. This involves new approaches to ensure the predictability and integrity of AI-driven decisions, a topic that will be central to the evolution of all avionics standards moving forward.
