Artificial noise elimination (ANE) has arisen as a promising strategy to counteract artificial noise (AN) at the eavesdropper (Eve). However, conventional ANE schemes rely on the attainable channel state information (CSI) between the transmitter (Alice) and legitimate receiver (Bob), which may reduce the feasibility of ANE. To break the gridlock, we focus on the ANE without the Alice-Bob link CSI by minimizing the artificial-noise-to-signal ratio (ANSR). Specifically, the detailed minor component analysis (MCA)-based ANE scheme is proposed to eliminate the AN for a specific classification. Moreover, the principal component analysis (PCA)-based ANE scheme is proposed for the case of binary phase shift keying (BPSK). Furthermore, the k-nearest neighbor (kNN)-based ANE scheme is proposed to exploit the additional information from the unclassified samples. To evaluate the cost of computing, the computational complexity of the aforementioned ANE schemes is analyzed. Meanwhile, the minimum required numbers of observed vectors for the MCA and PCA are derived. Finally, our numerical simulations show that proposed ANE schemes outperform the conventional hyperplane clustering (HC) scheme and can achieve the ANSR of ${\mathbf {1}}{{\mathbf {0}}^{{\mathbf { - 4}}}}$ at the signal-to-noise ratio (SNR) of 25 dB.