Distributed precoding has proven to be capable of enhancing the secrecy capacity of the multi-relay wiretap system. An iterative distributed precoding and channel state information (CSI) sharing scheme can be used to reduce the CSI overhead at each relay node. However, in practical applications, the CSI of each relay node cannot be perfectly known to themselves, especially that of the relay-eavesdropper channels. Thus, partial CSI for the relay-eavesdropper links is assumed, and the corresponding distributed precoding and CSI sharing schemes are investigated. Under the assumption that the average value of the relay-eavesdropper channel is known at each relay node, an extended iterative distributed precoding and CSI sharing scheme is proposed. Simulation results demonstrate that with the increase of the power ratio of the constant part to the random part of the relay-eavesdropper channels, the proposed scheme with partial CSI performs increasingly close to the one with perfect CSI in secrecy capacity. Figure 2. Secrecy capacity of the multi-relay system with M D 2.the Rician channel than with the Rayleigh channel. The proposed ITP with the average channel values performs better with the increase of K. The gap between the ITP schemes with perfect and partial CSI narrows with the increase of K, because the random parts of the R i -E channels become smaller.The secrecy capacity results are plotted in Figure 3 with M D 3. The observation from Figure 3 is similar to that from Figure 2. The difference between ITP with perfect and partial CSI is negligible as can be seen from Figure 3(a). The relationship among the ITP schemes with different CSI assumptions can be clearly observed from Figure 3(c). When K is small and the R i -E channels are Figure 3. Secrecy capacity of the multi-relay system with M D 3.almost unknown, the ITP w/ AV scheme has similar performance as that without the R i -E CSI. If K is larger than 10 dB, the gap between ITP with perfect and partial CSI is negligible.The results in Figures 2(c) and 3(c) demonstrate that the proposed ITP schemes with partial CSI is suited for the Rician channel with the average channel values available at the relay nodes. The power ratio K between the constant and random parts affects the performance of the proposed scheme. As shown in Figures 2(c) and 3(c), if K < 4 dB, namely, each relay node almost knows nothing about the related R i -E channel, the ITP w/AV scheme performs close to that without the eavesdropper information. On the other hand, if K > 10 dB, the proposed scheme with partial CSI is similar to that with perfect CSI.