Microblog have become very popular in recent years. Sina Weibo is the top microblogging network in China. The popularity of Sina Weibo has attracted a large number of companies to do marketing on it. However, the Weibo marketing suffers from zombie followers, a kind of fake accounts. In this paper, ZDACF(zombie follower detection approach based on the number of common friend),a social graph based approach is proposed to detect zombie followers. ZDACF use the feature of social graph to detect zombie followers. ZDACF crawl Sina Weibo using Sina Weibo API to collect the follower dataset, and calculate the Nearest Neighbor Homology to distinguish the zombie followers from normal ones. We measure the effectiveness of ZDACF, the result of evaluation shows that the detection rate of ZDACF is better than the previously work on different classifiers respectively.