Existing steganalysis systems are usually assumed that their target steganography and configurations are known. However, this assumption does not hold in actual implementation. The adaptability of different steganalysis systems varies during the detection of different steganographic algorithms with different configurations. In this paper, we define such applicability as the blindness and create a specific evaluation criterion based on the probability model for steganographic systems, and analyse the way to improve the blindness including the features fusion and optimisation as well as the decision level fusion. Experimental results show that, our criterion is effective in indicating the adaptabilities of steganalysis system on specific steganographic system, and the proposed schemes do have raised the blindness.%现有隐写分析系统通常假定隐写算法及其配置已知,但该条件在实际的应用环境中很难满足.实际上,不同的隐写分析系统对未知隐写算法和配置有不同的适用性.将该适用性定义为盲性并在隐写系统概率模型的基础上给出一种具体的评价指标,分析提高盲性的方法,包括特征的融合优选和决策层的融合.实验表明,提出的指标能够有效指示隐写分析系统对具体隐写系统的适用性,所提方法确实提高了盲性.
展开▼
