ProPETUS: Robust Prediction-Driven Spectrum Allocation for CRN-IoT Under SSDF and PUEA

Prediction-driven spectrum allocation can improve channel access in Cognitive Radio Network-enabled Internet of Things (CRN-IoT) systems, but its reliability depends on the integrity of the sensing history. Spectrum Sensing Data Falsification (SSDF) corrupts cooperative sensing reports, while Primary User Emulation Attacks (PUEA) distort the observed occupancy process. This paper presents ProPETUS, a distributed proactive spectrum allocation framework for joint SSDF and PUEA conditions. ProPETUS combines lightweight truncated-SVD sensing reconstruction motivated by matrix-completion theory, leave-one-out trust-weighted fusion, EWMA-based short-horizon prediction and round-robin allocation. The main contribution lies in the decision-centric integration and matched evaluation of these lightweight components under joint-threat settings. Matched Monte Carlo simulations show improved prediction accuracy and RMSE relative to MV-CSS, ProBLeSS and ProMETHEUS under the evaluated operating conditions. The access-level results show lower severe-point backoff and switching-delay values, while channel utilization and fairness remain largely constrained by the available idle-spectrum opportunity budget. Energy results indicate that the prediction and access-stability gains are achieved without a substantial total-energy penalty.