Contemporary security measures are based on the concept of "defending with walls," and have been unable to escape the endless cycle in which attackers and defenders continuously strengthen their spears and shields. The anticipated emergence of hypothetical high-performance attack AI such as Claude Mythos has further exacerbated this structural limitation. This paper proposes RIS (Raphael Immune System), a new security architecture that fundamentally resolves this problem. RIS places the human body's immune system at the core of its design philosophy, shifting from "defending with walls" to "strengthening white blood cells and platelets." Specifically, it adopts a three-layer structure: a natural immunity layer in which white blood cells directly repel known threats; a blockade layer in which exits function as platelets, sealing intruders inside; and an acquired immunity layer in which unknown threats are detected by causing them to self-inflate through a "trial gate" trap structure. Unlike conventional attacks, which are completed externally using large-scale programs to breach walls, RIS requires attackers to construct exit programs internally after infiltration, structurally imposing the cost of inflation and self-exposure. This realizes a paradoxical structure in which attackers believe they have succeeded while becoming cornered.
信吾 板倉 (2026) studied this question.