Cognitive Radio (CR) networks enable dynamic spectrum access and can significantly improve spectral efficiency. Cooperative Spectrum Sensing (CSS) exploits the spatial diversity between CR users to increase sensing accuracy. However, in a realistic scenario, the trustworthy of CSS is vulnerable to Spectrum Sensing Data Falsification (SSDF) attack. In an SSDF attack, some malicious CR users deliberately report falsified local sensing results to a data collector or Fusion Center (FC) and, then, affect the global sensing decision. In the present study, we investigate an analytical model for a hard SSDF attack and propose a robust defense strategy against such an attack. We show that FC can apply learning and estimation methods to obtain the attack parameters and use a better defense strategy. We further assume a log-normal shadow fading wireless environment and discuss the attack parameters that can affect the strength of SSDF attack. Simulation results illustrate the effectiveness of the proposed defense method against SSDF attacks, especially when the malicious users are in the majority.