Personal Genomic information is becoming increasingly important to both scientific research and clinical practice. However, security breach, misuse, or unintended disclosure of this information may result in severe privacy breaches. Traditional privacy preservation of personal genome information is implemented in an "all-or-none" manner, i.e., an entire genome being controlled as either fully accessible or fully inaccessible. In this paper, we propose a new fine-grained privacy protection method for flexible multi-level genome information protection and access. The method can make use of any user-defined hierarchical knowledge structure to define privacy levels and control cryptography-based hierarchical access. It also implements dual-stage encryptions to allow efficient definition, addition, and update of feature-based privacy protections. The experiments show that it can be effectively implemented to deal with real personal genome data sets in the future.