隱寫術，這項古老而神秘的技術，早在西元1499年就被定義出來。它以一種巧妙的方式將秘密訊息隱藏在看似普通的載體之中，使得訊息的接收者能夠在不引起外界注意的情況下解讀出內容。藏頭詩是其中一種常見的隱寫術方法，但隨著科技的飛速進步，這項技術已經突飛猛進，呈現出更多元化的應用方式，包括音檔、影片、圖片等不同媒介，為隱寫術的研究與應用開辟了更廣闊的領域。特別是圖片這一媒介，其所帶來的影響力遠遠超越文字，因為影像的可視化特性使其更能夠引起人們的注意和共鳴。由於科技的發展，隱寫術的技術也隨之精進，現在只需對影像進行數學計算，就能將資訊巧妙地嵌入其中，為信息隱藏帶來更高的效率和可靠性。在圖像隱寫的技術上，主要分為頻域與空間域兩種方式，各自具有獨特的優缺點，而近年來深度學習的蓬勃發展，更為圖像隱寫開啟了新的可能性，使得應用範圍更加廣泛。本研究將採用傳統的頻域圖像隱寫方法，並融合離散小波卷積的深度學習模型，以期進一步提高圖像隱寫的不可感知性、隱密性以及嵌入的圖像大小等特性。值得一提的是，在實驗過程中，我們採用了多層的小波卷積，讓資訊可以被一層一層地嵌入其中。相較於傳統的單張圖像隱寫，多層小波卷積技術允許更多張圖像被隱藏在其中，這將為特定應用場景提供更大的彈性與豐富的應用可能性。最後，除了技術研究，本研究亦規劃了一個創新的商業模式，試圖通過資訊的巧妙藏入，達到推薦行銷的目標。然而，我們強調在利用隱寫術進行行銷活動時，必須遵循合法和道德的原則，確保不侵犯他人隱私權，並避免造成任何潛在的損害。

The concept of steganography was formulated as early as 1499, encompassing a carrier and a covert communication. To transmit classified information, the concealed message would be infused into the carrier, with hidden messages in poetry being a frequent example. However, with the rapid progression of technology and the increasing expediency of information interchange, an immeasurable number of messages are being conveyed at all times, and the types of steganography are becoming increasingly diverse, including the utilization of audio, video, and images. It is noteworthy to mention images in particular, as a single image can have a far-reaching impact surpassing that of a string of text, and images are highly manipulable. Thus, steganography techniques have grown even more sophisticated, and by performing mathematical computations on images, information can be embedded within them. Image steganography can be broadly categorized into the frequency domain and spatial domain, with each having its own strengths and limitations. In recent years, with the surge of deep learning, this study adopts traditional frequency domain image steganography and combines it with a deep learning model that employs discrete wavelet convolution to enhance the imperceptibility, confidentiality, and embedded image size of image steganography. Furthermore, this study employs multi-layer wavelet convolution in experiments to permit information to be embedded layer by layer, as opposed to only in a single image, which can enable more images to be hidden. Finally, this study also outlines a commercial model that can utilize the embedded information for the purpose of recommendation marketing.

中文摘要 vi
Abstract vii
目錄 viii
圖目錄 x
表目錄 xii
第一章 緒論 1
1.1 研究背景 1
1.2 研究動機與目的 1
第二章 文獻探討 4
2.1 資訊隱藏(Information Hiding) 4
2.2 圖像隱寫術(Image Steganography) 6
2.2.1 空間域圖像隱寫(Spatial domain Steganography) 7
2.2.2 頻率域圖像隱寫(Frequency domain Steganography) 7
2.3 小波轉換(Wavelet Transform) 8
2.3.1 離散小波轉換(Discrete Wavelet Transform，DWT) 8
2.4 深度學習(Deep learning，DL) 9
2.4.1 自動編碼器(AutoEncoder) 10
2.4.2 多層小波卷積(Multi-level Wavelet Transform CNN，MWCNN) 11
2.4.3 卷積神經網路(CNN) 13
2.5 隱寫圖像的品質衡量 13
2.5.1 峰值訊噪比(Peak signal-to-noise ratio, PSNR) 13
2.5.2 結構相似性(Structural similarity, SSIM) 15
2.6 推薦行銷(Referral Marketing) 16
2.7 與深度圖像隱寫的相關研究 17
第三章 研究方法 18
3.1 資料集 18
3.2 U-net隱寫術 19
3.3 離散小波卷積單層圖像隱寫 19
3.3.1 損失函數 21
3.4 離散小波卷積多層圖像隱寫 22
3.5 藉由圖像隱寫進行推薦行銷 24
第四章 實驗結果 26
4.1 實驗環境 26
4.2 實驗過程與結果 26
4.2.1 單層單張圖像隱寫 26
4.2.2 雙層單張圖像隱寫 30
4.2.3 多層單張圖像隱寫 33
第五章 結論與未來展望 39
參考文獻 42

[1]W. Bender, D. Gruhl, N. Morimoto, and A. Lu.(1996)Techniques for data hiding, IBM syst. J., vol. 35, nos. 3 – 4, pp. 313 – 336.
[2]Nandhini Subramanian, Omar Elharrouss, Somaya Al-Maadeed, and Ahmed Bouridane.(2021) Image Steganography: A Review of the Recent Advances. IEEE Digital Object Identifier 10.1109
[3]Shumeet Baluja.(2017)Hiding Images in Plain Sight: Deep Steganography. 31st Conference on Neural Information Processing Systems
[4]D. Hu, L. Wang, W. Jiang, S. Zheng and B. Li, "A Novel Image Steganography Method via Deep Convolutional Generative Adversarial Networks," in IEEE Access, vol. 6, pp. 38303-38314, 2018.
[5]Kweon, H., Park, J., Woo, S., & Cho, D. (2021). Deep multi-image steganography with private keys. Electronics, 10(16), 1906.
[6]P.J.Liu, H.Z.Zhang, K.Zhang, L.Lin, and Wang M.Z.Wang.(2019) .Multi-level Wavelet-CNN for Image Restoration
[7]Petitcolas, F. A. P., Anderson, R. J., & Kuhn, M. G. (1999). Information hiding-A survey. Proceedings of the IEEE, 87(7), 1062–1078.
[8]Trithemius, J. (1606). Steganographia, hoc est, Ars per occvltam scriptvram animi svi volvntatem absentibvs aperiendi certa ...: Præfixa Est Hvic operi Sva clavis, Sev Vera Introductio ab ipso Authore Concinnata: Hactenvs qvidem a mvltis Mvltvm Desiderata, Sed À paucissimis visa: Nunc vero in Gratiam secretioris Philosophiæ studiosorum publici iuris facta. Ex officina typographica Matthiæ Beckeri, sumptibus Ioannis Berneri.
[9]Petitcolas, F. A. P., Anderson, R. J., & Kuhn, M. G. (1999). Information hiding-A survey. Proceedings of the IEEE, 87(7), 1062–1078.
[10]Hussain, M., Wahab, A. W., Idris, Y. I., Ho, A. T. S., & Jung, K.-H. (2018). Image steganography in spatial domain: A survey. Signal Processing: Image Communication, 65, 46–66.
[11]Banoci, V., Bugar, G., & Levicky, D. (2011). A novel method of image steganography in DWT domain. Proceedings of 21st International Conference Radioelektronika 2011.
[12]V, S. (2020). An empirical science research on bioinformatics in Machine Learning. JOURNAL OF MECHANICS OF CONTINUA AND MATHEMATICAL SCIENCES, spl7(1).
[13]Michelucci, U. (2022). Autoencoders. Applied Deep Learning with TensorFlow 2, 257–283.
[14]Ronneberger, O. (2017). Invited talk: U-net convolutional networks for biomedical image segmentation. Informatik Aktuell, 3–3.
[15]Duan, X., Jia, K., Li, B., Guo, D., Zhang, E., & Qin, C. (2019). Reversible image steganography scheme based on a U-Net Structure. IEEE Access, 7, 9314–9323.
[16]Mallat, S. G. (1989). A theory for multiresolution signal decomposition: The wavelet representation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 11(7), 674–693.
[17]Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278–2324.
[18]Wani, M. A., & Sultan, B. (2022). Deep Learning Based Image Steganography: A Review. WIREs Data Mining and Knowledge Discovery.
[19]Baluja, S. (2017). Hiding images in plain sight: Deep steganography. Advances in neural information processing systems, 30.
[20]Volkhonskiy, D., Nazarov, I., &; Burnaev, E. (2020). Steganographic generative Adversarial Networks. Twelfth International Conference on Machine Vision (ICMV 2019).
[21]Shi, H., Dong, J., Wang, W., Qian, Y., & Zhang, X. (2018). SSGAN: Secure steganography based on generative Adversarial Networks. Advances in Multimedia Information Processing – PCM 2017, 534–544.
[22]Duan, X., Li, B., Guo, D., Zhang, Z., & Ma, Y. (2020). A coverless steganography method based on generative Adversarial Network. EURASIP Journal on Image and Video Processing, 2020(1).
[23]Van, T. P., Dinh, T. H., & Thanh, T. M. (2019). Simultaneous convolutional neural network for highly efficient image steganography. 2019 19th International Symposium on Communications and Information Technologies (ISCIT).
[24]ur Rehman, A., Rahim, R., Nadeem, S., & ul Hussain, S. (2019). End-to-end trained CNN encoder-decoder networks for image steganography. Lecture Notes in Computer Science, 723–729.
[25]Strikingly. (n.d.). 什麼是推薦行銷? - 建立你的網站. Strikingly. Retrieved March 9, 2023, from https://tw.strikingly.com/content/blog/referral-marketing/
[26]Referral marketing: What is it and why does it matter? (n.d.). Retrieved March 13, 2023, from https://www.business2community.com/marketing/referral-marketing-matter-0944226
[27]Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., Berg, A. C., & Fei-Fei, L. (2015). Imagenet Large Scale Visual Recognition Challenge. International Journal of Computer Vision, 115(3), 211–252.
[28]Demir, I., Koperski, K., Lindenbaum, D., Pang, G., Huang, J., Basu, S., Hughes, F., Tuia, D., & Raskar, R. (2018). DeepGlobe 2018: A Challenge to parse the Earth through satellite images. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW).
[29]Das, A., Wahi, J. S., Anand, M., & Rana, Y. (2021). Multi-Image Steganography Using Deep Neural Networks. ArXiv Computer Vision and Pattern Recognition
[30]ur Rehman, A., Rahim, R., Nadeem, S., & ul;Hussain, S. (2019). End-to-end trained CNN encoder-decoder networks for image steganography. Lecture Notes in Computer Science, 723–729.
[31]Duan, X., Liu, N., Gou, M., Wang, W., & Qin, C. (2020). STEGANOCNN: Image steganography with generalization ability based on convolutional neural network. Entropy, 22(10), 1140.
[32]Zhu, J., Kaplan, R., Johnson, J., & Fei-Fei, L. (2018). Hidden: Hiding data with Deep Networks. Computer Vision – ECCV 2018, 682–697.
[33]Fu, Z., Wang, F., & Cheng, X. (2020). The secure steganography for hiding images via gan. EURASIP Journal on Image and Video Processing, 2020(1).