Decoding-Order-Aware Secrecy Rate Maximization in IRS-Assisted Untrusted NOMA Under Imperfect SIC

Abstract

This work investigates an IRS-assisted downlink NOMA system with untrusted users under imperfect SIC, where internal eavesdropping-users intercepting each other's messages-poses distinct security challenges absent in conventional external-eavesdropper models. A key contribution is the systematic analysis of all possible decoding orders, proving that the (2,2) order yields the largest feasible power allocation region satisfying simultaneous secrecy and QoS constraints. The joint optimization of decoding order, power allocation, and IRS reflection coefficients is decomposed into two tractable subproblems: closed-form power allocation under perfect SIC and semidefinite relaxation (SDR)-based IRS phase optimization, solved via an alternating framework. Numerical results demonstrate performance improvements of 14.6% at low power (10 W) and 12.7% at high power (80 W) over benchmark schemes, validating the effectiveness of the proposed approach in balancing secrecy and service quality under practical SIC imperfections.