Physical layer security in the presence of passive eavesdroppers (i.e., the eavesdroppers' channels are unknown to the transmitter) in correlated channels is addressed. The majority of the literature with passive eavesdroppers assumption focus on uncorrelated channels and random artificial noise (AN) generation in the null space of the legitimate channel. This paper targets joint optimization of the AN vector and the power fraction used for beamforming in correlated channels. These two variables are obtained by maximizing the minimum secure probability of attaining a pre-determined secrecy level for a certain number of the most dangerous potential Eves, which are obtained by the correlation map. The correlation map contains long-term channel correlation matrices for all the potential Eves' locations, which are captured by legitimate users passing through these areas in the past. We use a secure probability map to illustrate the secrecy levels for all the possible positions of the legitimate user. The results show the significant secrecy enhancement of our proposed algorithm when compared to the exiting AN generation methods.