In this paper, a conventional FeFET-based 3T2R neuron with tunable leaky effect, enabling both excitatory and inhibitory input connections is proposed. The 3T2R neuron is only composed of three transistors and two resistors, while can realize the full LIF function without an additional customization process. A SNN designed for the Sudoku task based on 3T2R neurons is efficiently implemented with lower hardware cost, and due to the conveniently tunable leaky effect, compared with the counterpart whose leakage effect cannot be changed, the FoM, which quantifies the competitive advantage of the winner neuron is improved by 30% on average.