Resilient network design for cognitive radio networks under spectrum sensing data falsification attacks

Abstract

Cognitive radio networks (CRNs) rely on cooperative spectrum sensing (CSS), but fusing reports from distributed secondary users (SUs) exposes the network to Spectrum Sensing Data Falsification (SSDF) attacks. Most countermeasures evaluate SSDF defenses at the sensing layer and leave open whether the network still meets its service requirements under attack. This paper presents SC-Joint, an integration framework that treats SSDF defense as an admission constraint within a cross-layer problem. A Beta–Bernoulli reputation layer admits SUs, and the cooperating set and sensing time are then jointly selected subject to QoS-SLA constraints on detection probability, throughput and network lifetime. The reputation is oracle-free: each SU is scored by its agreement with the leave-one-out majority of the other SUs, never against ground-truth labels. Under OR fusion, a single always-yes attacker saturates the global false-alarm probability and drives throughput to zero, and under nested injection the undefended baseline loses feasibility as attackers are added (Propositions 1–2). We derive an admission condition (Proposition 3) that predicts the falsification probability below which stealthy attackers are admitted (p* ≈ 0.44) and the PU activity above which honest high-SNR SUs are excluded (P(H₁) ≈ 0.42). MATLAB simulations of a 12-SU network show that undefended OR fusion loses all feasibility once one attacker is admitted, whereas SC-Joint sustains 79.7–100% feasible allocation at about 2.1 Mb/s, and 97–100% under network stress, matching oracle exclusion in paired tests. The ablation attributes the gain to trust-based admission, while sensing-time optimization mainly secures lifetime. The mechanism is limited to OR fusion, P(H₁) < 0.45 and non-colluding attackers, and stealthy attackers with p < p* are admitted.