In recent days, the advent of 5G/6G networks demands the robust and secure communication protocols. However, the traditional channel estimation methods are vulnerable to the security threats. In this research, the least square estimation with terrestrial unmanned aerial vehicles (LSE-UAV) is proposed for secure channel estimation. Initially, the multiple input multiple output (MIMO) system is designed to ensure the reliability. Then, the data transmission rates are further enhanced with the help of multiple antennas at the transmitter and receiver. After that, a baseline for assessing channel estimation technique is established with wideband MIMO orthogonal frequency division multiple access (WMIMO-OFDMA) model. Further, OFDMA is employed to categorize the bandwidth into multiple orthogonal subcarriers. Next, LSE approach is employed to estimate channel impulse response (CIR) and channel frequency response (CFR). From the results, proposed model obtained high results in the context of root mean square error (RMSE) with 0.0123 dB, mean absolute percentage error (MAPE) with 0.9201 dB, mean absolute error (MAE) with 0.0134 dB, and coefficient of determination (R2) with 7.7653 dB when compared to existing convolutional neural network and long short-term memory (CNN-LSTM).