Identifying the aim of a pursuer quickly and accurately is vital to taking appropriate measures and successfully avoiding capture. In order to determine whether an evader in a flying formation is aimed by the pursuer or not, an aim identification algorithm is proposed based on the stationarity difference of the mixed product of three direction vectors: 1) the evader's velocity vector; 2) the evader-to-pursuer line-of-sight (LOS) vector; and 3) at the same instant another such LOS vector at a subsequent instant. Due to the effect of the guidance law, the subsequent LOS always shows a tendency to keep coplanar with the velocity of the aim and the previous LOS such that the mixed product obtained by the aim changes very slowly. In other words, the mixed product sequence obtained by the aim based on the known velocity and the bearing measurements of the LOS has relatively good stationarity, while the stationarity of the sequence obtained by the evader not being aimed at is poor. Based on the stationarity difference, the mixed product sequences are used to formulate a chi-square distributed statistic such that the aim identification is equivalently transformed into a hypothesis testing problem. Simulation results are given to verify the effectiveness of the proposed algorithm.