CAJ | 학술논문

现有盲检测技术在实际检测中，由于嵌入算法未知导致检测困难。为此，提出一种基于Boosting算法融合的图像隐写分析方法。通过训练分类器建立不同隐写算法下的分类器模型，利用 Boosting算法计算各分类器的分类性能，对各分类器的概率输出进行融合，得到最终检测结果。基于典型空间域隐写算法和JPEG隐写算法的实验结果表明，该方法实现了对多种隐写算法的有效检测，应用Boosting算法融合后整体检测性能提升了约2%。
현유맹검측기술재실제검측중，유우감입산법미지도치검측곤난。위차，제출일충기우Boosting산법융합적도상은사분석방법。통과훈련분류기건립불동은사산법하적분류기모형，이용 Boosting산법계산각분류기적분류성능，대각분류기적개솔수출진행융합，득도최종검측결과。기우전형공간역은사산법화JPEG은사산법적실험결과표명，해방법실현료대다충은사산법적유효검측，응용Boosting산법융합후정체검측성능제승료약2%。
The existing detection algorithms are difficult to obtain high detection accuracy when applied to the condition, in which the embedding algorithm of the stego-images is unknown. Therefore, this paper proposes a steganography-unknown image steganalysis method based on Boosting fusion. It obtains various classifying results by establishing steganography algorithm classifier models in the training phase, and acquires the performance of these classifies according to the Boosting algorithm. The final detection result is obtained by combinational rule based on probability output. The detection work is presented to attack the current different spatial domain and JPEG steganographic algorithms. Extensive experimental results show that this proposed method is effective for multi-steganographic algorithms, and Boosting takes advantage of the individual strengths from each detection system and whole detection performance is probably increased by 2%.