Reliability allocation aims to distribute system-level performance requirements across components or barrier functions while satisfying safety acceptance criteria. At early design stages, when system realisations are not yet defined, fault tree models describe functional relationships between initiating events, barriers, and consequences. In this setting, component-level cost information is generally unavailable, and reliability serves as a proxy for implementation effort under an assumed monotonic relationship. The allocation problem is inherently underdetermined as many combinations of function reliabilities can satisfy a given top-level constraint. A principled approach to selecting among these feasible solutions can therefore provide a risk-informed basis for design decisions.

We present a multi-objective optimisation method in which the feasible set is rendered convex through a logarithmic reparametrisation of cutset contributions. Convex surrogates for entropy-like objectives are constructed, enabling the systematic exploration of trade-offs between risk distribution across cutsets and reliability dispersion across barrier functions, generating structured families of optimal solutions.