Pilot contamination attack is an important kind of active eavesdropping activity conducted by a malicious user during channel training phase. In this paper, motivated by the fact that frequency asynchronism could introduce divergence of the transmitted pilot signals between intended user and attacker, we propose a new uncoordinated frequency shift (UFS) scheme for detection of pilot contamination attack in multiple antenna system. An attack detection algorithm is further developed based on source enumeration method. Both the asymptotic detection performance analysis and numerical results are provided to verify the proposed studies. The results demonstrate that the proposed UFS scheme can achieve comparable detection performance as the existing superimposed random sequence based scheme, without sacrifice of legitimate channel estimation performance.