Despite of massive progress in drug discovery field, cancer incidence and mortality rates have not decreased yet. In this situation, in-silico based drug repositioning is getting a spotlight because of relatively slow development in conventional anticancer drug discovery field. In this research, we predicted drugs that show new anticancer effects different from known original effects. To do this, we constructed protein-protein interaction networks that are perturbed by mutated proteins and find neighbors of mutated ones which reflect cancer states. We prioritized drugs using a score function and searched literature evidences for our repositioned drug candidates. In result, performance of drug prediction (AUROC with ranked answer drugs) is better than previous works. In addition, we propose new drug targets that are more important in cancer and not in normal using network propagation algorithm. Through this, we expect that we can find drugs with higher sensitivity and lower resistance.