Welcome to the 5th ACM Workshop on Information Hiding and Multimedia Security Workshop -- IH&MMSec'17 in Philadelphia, PA, held June 20-21, 2017. In response to our call for papers, 34 excellent papers were submitted from authors throughout North America, Europe, and Asia. The best 18 of these papers were accepted (53% acceptance rate) and assembled into a strong technical program. The accepted papers cover the fields of steganography and steganalysis in digital media, multimedia forensics, digital watermarking, data hiding in natural language, deep learning approaches to both forensics and steganalysis. We sincerely thank all the submitting authors for their contributions, and the reviewers for their invaluable help. We expect the selected papers to be of wide interest to researchers working in the field and to participants from industry and from government institutions. The technical program also includes two invited keynote speakers. The first presentation is given by Dr. Anupam Das from Carnegie Mellon University on the topic of using motion sensors in smartphones to track users. The second presentation is given by Dr. Rachel Greenstadt from Drexel University on the topic of how stylometry and machine learning can be used to determine the author of both written documents and software. As usual, the workshop is structured into three days with the afternoon of the second day devoted to a social event. The social event is designed to promote discussions and to help establish relationships for future collaboration among participants. Also, at the end of the second day before the start of the social event, time is reserved for a rump session during which the participants are encouraged to share their work in progress, discuss unpublished results, demo new products, and make relevant announcements. A great team effort put together the technical program. The Program Committee assisted by 29 external reviewers provided timely and high-quality reviews. A double-blind review process was used to ensure fairness. Each paper was carefully read and appraised by at least three reviewers, however the majority of papers were reviewed by four reviewers. To let the Program Chairs select the best quality and relevant work, papers with conflicting reviews were discussed at length. We thank all participants for their help in putting together this great program.
{"title":"Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security","authors":"M. Stamm, Matthias Kirchner, S. Voloshynovskiy","doi":"10.1145/3082031","DOIUrl":"https://doi.org/10.1145/3082031","url":null,"abstract":"Welcome to the 5th ACM Workshop on Information Hiding and Multimedia Security Workshop -- IH&MMSec'17 in Philadelphia, PA, held June 20-21, 2017. \u0000 \u0000In response to our call for papers, 34 excellent papers were submitted from authors throughout North America, Europe, and Asia. The best 18 of these papers were accepted (53% acceptance rate) and assembled into a strong technical program. The accepted papers cover the fields of steganography and steganalysis in digital media, multimedia forensics, digital watermarking, data hiding in natural language, deep learning approaches to both forensics and steganalysis. \u0000 \u0000We sincerely thank all the submitting authors for their contributions, and the reviewers for their invaluable help. We expect the selected papers to be of wide interest to researchers working in the field and to participants from industry and from government institutions. \u0000 \u0000The technical program also includes two invited keynote speakers. The first presentation is given by Dr. Anupam Das from Carnegie Mellon University on the topic of using motion sensors in smartphones to track users. The second presentation is given by Dr. Rachel Greenstadt from Drexel University on the topic of how stylometry and machine learning can be used to determine the author of both written documents and software. \u0000 \u0000As usual, the workshop is structured into three days with the afternoon of the second day devoted to a social event. The social event is designed to promote discussions and to help establish relationships for future collaboration among participants. Also, at the end of the second day before the start of the social event, time is reserved for a rump session during which the participants are encouraged to share their work in progress, discuss unpublished results, demo new products, and make relevant announcements. \u0000 \u0000A great team effort put together the technical program. The Program Committee assisted by 29 external reviewers provided timely and high-quality reviews. A double-blind review process was used to ensure fairness. Each paper was carefully read and appraised by at least three reviewers, however the majority of papers were reviewed by four reviewers. To let the Program Chairs select the best quality and relevant work, papers with conflicting reviews were discussed at length. We thank all participants for their help in putting together this great program.","PeriodicalId":431672,"journal":{"name":"Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security","volume":"141 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116459453","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this talk, I will discuss my lab's work in the emerging field of adversarial stylometry and machine learning. Machine learning algorithms are increasingly being used in security and privacy domains, in areas that go beyond intrusion or spam detection. For example, in digital forensics, questions often arise about the authors of documents: their identity, demographic background, and whether they can be linked to other documents. The field of stylometry uses linguistic features and machine learning techniques to answer these questions. We have applied stylometry to difficult domains such as underground hacker forums, open source projects (code), and tweets. I will discuss our Doppelgnger Finder algorithm, which enables us to group Sybil accounts on underground forums and detect blogs from Twitter feeds and reddit comments. In addition, I will discuss our work attributing unknown source code and binaries.
{"title":"Using Stylometry to Attribute Programmers and Writers","authors":"R. Greenstadt","doi":"10.1145/3082031.3092567","DOIUrl":"https://doi.org/10.1145/3082031.3092567","url":null,"abstract":"In this talk, I will discuss my lab's work in the emerging field of adversarial stylometry and machine learning. Machine learning algorithms are increasingly being used in security and privacy domains, in areas that go beyond intrusion or spam detection. For example, in digital forensics, questions often arise about the authors of documents: their identity, demographic background, and whether they can be linked to other documents. The field of stylometry uses linguistic features and machine learning techniques to answer these questions. We have applied stylometry to difficult domains such as underground hacker forums, open source projects (code), and tweets. I will discuss our Doppelgnger Finder algorithm, which enables us to group Sybil accounts on underground forums and detect blogs from Twitter feeds and reddit comments. In addition, I will discuss our work attributing unknown source code and binaries.","PeriodicalId":431672,"journal":{"name":"Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128515399","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
There are two interpretations of the term "square root law of steganography". As a rule of thumb, that the secure capacity of an imperfect stegosystem scales only with the square root of the cover size (not linearly as for perfect stegosystems), it acts as a robust guide in multiple steganographic domains. As a mathematical theorem, it is unfortunately limited to artificial models of covers that are a long way from real digital media objects: independent pixels or first-order stationary Markov chains. It is also limited to models of embedding where the changes are uniformly distributed and, for the most part, independent. This paper brings the theoretical square root law closer to the practice of digital media steganography, by extending it to cases where the covers are Markov Random Fields, including inhomogeneous Markov chains and Ising models. New proof techniques are required. We also consider what a square root law should say about adaptive embedding, where the changes are not uniformly located, and state a conjecture.
{"title":"The Square Root Law of Steganography: Bringing Theory Closer to Practice","authors":"Andrew D. Ker","doi":"10.1145/3082031.3083235","DOIUrl":"https://doi.org/10.1145/3082031.3083235","url":null,"abstract":"There are two interpretations of the term \"square root law of steganography\". As a rule of thumb, that the secure capacity of an imperfect stegosystem scales only with the square root of the cover size (not linearly as for perfect stegosystems), it acts as a robust guide in multiple steganographic domains. As a mathematical theorem, it is unfortunately limited to artificial models of covers that are a long way from real digital media objects: independent pixels or first-order stationary Markov chains. It is also limited to models of embedding where the changes are uniformly distributed and, for the most part, independent. This paper brings the theoretical square root law closer to the practice of digital media steganography, by extending it to cases where the covers are Markov Random Fields, including inhomogeneous Markov chains and Ising models. New proof techniques are required. We also consider what a square root law should say about adaptive embedding, where the changes are not uniformly located, and state a conjecture.","PeriodicalId":431672,"journal":{"name":"Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125053712","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Watermarking natural language is still a challenge in the domain of digital watermarking. Here, only the textual information must be used as a cover. No format changes or modified illustrations are accepted. Still, natural language watermarking (NLW) has some important applications, especially in leakage tracking, where a small set of individually marked copies of a confidently text is distributed. Properties of watermarking schemes such as imperceptibility, blindness or adaptability to non-English languages are of importance here. In order to address these three simultaneously, we present a blind NLW scheme, consisting of four independent embedding methods, which operate on the phonetical, morphological, lexical and syntactical layer of German texts. An evaluation based on 1,645 assessments provided by 131 test persons reveals promising results.
{"title":"Towards Imperceptible Natural Language Watermarking for German","authors":"Oren Halvani, M. Steinebach, L. Graner","doi":"10.1145/3082031.3084682","DOIUrl":"https://doi.org/10.1145/3082031.3084682","url":null,"abstract":"Watermarking natural language is still a challenge in the domain of digital watermarking. Here, only the textual information must be used as a cover. No format changes or modified illustrations are accepted. Still, natural language watermarking (NLW) has some important applications, especially in leakage tracking, where a small set of individually marked copies of a confidently text is distributed. Properties of watermarking schemes such as imperceptibility, blindness or adaptability to non-English languages are of importance here. In order to address these three simultaneously, we present a blind NLW scheme, consisting of four independent embedding methods, which operate on the phonetical, morphological, lexical and syntactical layer of German texts. An evaluation based on 1,645 assessments provided by 131 test persons reveals promising results.","PeriodicalId":431672,"journal":{"name":"Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security","volume":"74 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132855490","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this paper, a new multi-attribute and high capacity image-based watermarking technique for relational data is proposed. The embedding process causes low distortion into the data considering the usability restrictions defined over the marked relation. The conducted experiments show the high resilience of the proposed technique against tuple deletion and tuple addition attacks. An interesting trend of the extracted watermark is analyzed when, within certain limits, if the number of embedded marks is small, the watermark signal far from being compromised, discretely improves in the case of tuple addition attacks. According to the results, marking 13% of the attributes and under an attack of 100% of tuples addition, 96% of the watermark is extracted. Also, while previous techniques embed up to 61% of the watermark, under the same conditions, we guarantee to embed 99.96% of the marks.
{"title":"A Minimum Distortion: High Capacity Watermarking Technique for Relational Data","authors":"M. L. P. Gort, C. F. Uribe, J. Nummenmaa","doi":"10.1145/3082031.3083241","DOIUrl":"https://doi.org/10.1145/3082031.3083241","url":null,"abstract":"In this paper, a new multi-attribute and high capacity image-based watermarking technique for relational data is proposed. The embedding process causes low distortion into the data considering the usability restrictions defined over the marked relation. The conducted experiments show the high resilience of the proposed technique against tuple deletion and tuple addition attacks. An interesting trend of the extracted watermark is analyzed when, within certain limits, if the number of embedded marks is small, the watermark signal far from being compromised, discretely improves in the case of tuple addition attacks. According to the results, marking 13% of the attributes and under an attack of 100% of tuples addition, 96% of the watermark is extracted. Also, while previous techniques embed up to 61% of the watermark, under the same conditions, we guarantee to embed 99.96% of the marks.","PeriodicalId":431672,"journal":{"name":"Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133191890","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yifeng Zhan, Yifang Chen, Qiong Zhang, Xiangui Kang
There have been a growing number of interests in using the convolutional neural network(CNN) in image forensics, where some excellent methods have been proposed. Training the randomly initialized model from scratch needs a big amount of training data and computational time. To solve this issue, we present a new method of training an image forensic model using prior knowledge transferred from the existing steganalysis model. We also find out that CNN models tend to show poor performance when tested on a different database. With knowledge transfer, we are able to easily train an excellent model for a new database with a small amount of training data from the new database. Performance of our models are evaluated on Bossbase and BOW by detecting five forensic types, including median filtering, resampling, JPEG compression, contrast enhancement and additive Gaussian noise. Through a series of experiments, we demonstrate that our proposed method is very effective in two scenario mentioned above, and our method based on transfer learning can greatly accelerate the convergence of CNN model. The results of these experiments show that our proposed method can detect five different manipulations with an average accuracy of 97.36%.
{"title":"Image Forensics Based on Transfer Learning and Convolutional Neural Network","authors":"Yifeng Zhan, Yifang Chen, Qiong Zhang, Xiangui Kang","doi":"10.1145/3082031.3083250","DOIUrl":"https://doi.org/10.1145/3082031.3083250","url":null,"abstract":"There have been a growing number of interests in using the convolutional neural network(CNN) in image forensics, where some excellent methods have been proposed. Training the randomly initialized model from scratch needs a big amount of training data and computational time. To solve this issue, we present a new method of training an image forensic model using prior knowledge transferred from the existing steganalysis model. We also find out that CNN models tend to show poor performance when tested on a different database. With knowledge transfer, we are able to easily train an excellent model for a new database with a small amount of training data from the new database. Performance of our models are evaluated on Bossbase and BOW by detecting five forensic types, including median filtering, resampling, JPEG compression, contrast enhancement and additive Gaussian noise. Through a series of experiments, we demonstrate that our proposed method is very effective in two scenario mentioned above, and our method based on transfer learning can greatly accelerate the convergence of CNN model. The results of these experiments show that our proposed method can detect five different manipulations with an average accuracy of 97.36%.","PeriodicalId":431672,"journal":{"name":"Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129512850","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Detection of modern JPEG steganographic algorithms has traditionally relied on features aware of the JPEG phase. In this paper, we port JPEG-phase awareness into the architecture of a convolutional neural network to boost the detection accuracy of such detectors. Another innovative concept introduced into the detector is the "catalyst kernel" that, together with traditional high-pass filters used to pre-process images allows the network to learn kernels more relevant for detection of stego signal introduced by JPEG steganography. Experiments with J-UNIWARD and UED-JC embedding algorithms are used to demonstrate the merit of the proposed design.
{"title":"JPEG-Phase-Aware Convolutional Neural Network for Steganalysis of JPEG Images","authors":"Mo Chen, V. Sedighi, M. Boroumand, J. Fridrich","doi":"10.1145/3082031.3083248","DOIUrl":"https://doi.org/10.1145/3082031.3083248","url":null,"abstract":"Detection of modern JPEG steganographic algorithms has traditionally relied on features aware of the JPEG phase. In this paper, we port JPEG-phase awareness into the architecture of a convolutional neural network to boost the detection accuracy of such detectors. Another innovative concept introduced into the detector is the \"catalyst kernel\" that, together with traditional high-pass filters used to pre-process images allows the network to learn kernels more relevant for detection of stego signal introduced by JPEG steganography. Experiments with J-UNIWARD and UED-JC embedding algorithms are used to demonstrate the merit of the proposed design.","PeriodicalId":431672,"journal":{"name":"Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131948545","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We propose a novel text steganography method using RNN Encoder-Decoder structure to generate quatrains, one genre of Chinese poetry. Compared to other text-generation based steganography methods which have either very low embedding rate or flaws in the naturalness of generated texts, our method has higher embedding rate and better text quality. In this paper, we use the LSTM Encoder-Decoder model to generate the first line of a quatrain with a keyword and then generate the following lines one by one. RNN has proved effective in generating poetry, but when applied to steganograpy, poetry quality decreases sharply, because of the redundancy we create to hide information. To overcome this problem, we propose a template-constrained generation method and develop a word-choosing approach using inner-word mutual information. Through a series of experiments, it is proven that our approach outperforms other poetry steganography methods in both embedding rate and poetry quality.
{"title":"Text Steganography with High Embedding Rate: Using Recurrent Neural Networks to Generate Chinese Classic Poetry","authors":"Yubo Luo, Yongfeng Huang","doi":"10.1145/3082031.3083240","DOIUrl":"https://doi.org/10.1145/3082031.3083240","url":null,"abstract":"We propose a novel text steganography method using RNN Encoder-Decoder structure to generate quatrains, one genre of Chinese poetry. Compared to other text-generation based steganography methods which have either very low embedding rate or flaws in the naturalness of generated texts, our method has higher embedding rate and better text quality. In this paper, we use the LSTM Encoder-Decoder model to generate the first line of a quatrain with a keyword and then generate the following lines one by one. RNN has proved effective in generating poetry, but when applied to steganograpy, poetry quality decreases sharply, because of the redundancy we create to hide information. To overcome this problem, we propose a template-constrained generation method and develop a word-choosing approach using inner-word mutual information. Through a series of experiments, it is proven that our approach outperforms other poetry steganography methods in both embedding rate and poetry quality.","PeriodicalId":431672,"journal":{"name":"Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116592688","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Alejandra Menendez-Ortiz, C. F. Uribe, José Juan García-Hernández
Reversible watermarking schemes (RWS) allow the restoration of the original signals after the watermarks are extracted. Most RWS for audio signals use time-domain for information hiding, although their transparency is hard to maintain for high embedding capacities. Some audio RWS use the frequency domain to improve transparency; however, their embedding capacity is lower than that of time-domain schemes. In this manuscript a RWS for audio signals is proposed, it differs from other schemes that work with the intDCT domain in the use of auditory masking properties, which are exploited to improve transparency, and the increase on embedding capacity is explored through a modified prediction error expansion (PEE). The payload capacity is 27.5 kbps with a degradation over -2 ODG, which are adequate results for practical audio applications. A generalized multi-bit expansion is proposed and experimental results suggest that higher expansion factors improve transparency.
{"title":"Audio Reversible Watermarking Scheme in the intDCT Domain with Modified Prediction Error Expansion","authors":"Alejandra Menendez-Ortiz, C. F. Uribe, José Juan García-Hernández","doi":"10.1145/3082031.3083246","DOIUrl":"https://doi.org/10.1145/3082031.3083246","url":null,"abstract":"Reversible watermarking schemes (RWS) allow the restoration of the original signals after the watermarks are extracted. Most RWS for audio signals use time-domain for information hiding, although their transparency is hard to maintain for high embedding capacities. Some audio RWS use the frequency domain to improve transparency; however, their embedding capacity is lower than that of time-domain schemes. In this manuscript a RWS for audio signals is proposed, it differs from other schemes that work with the intDCT domain in the use of auditory masking properties, which are exploited to improve transparency, and the increase on embedding capacity is explored through a modified prediction error expansion (PEE). The payload capacity is 27.5 kbps with a degradation over -2 ODG, which are adequate results for practical audio applications. A generalized multi-bit expansion is proposed and experimental results suggest that higher expansion factors improve transparency.","PeriodicalId":431672,"journal":{"name":"Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security","volume":"78 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129353219","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Chao Xia, Qingxiao Guan, Xianfeng Zhao, Zhoujun Xu, Yi Ma
The GFR (Gabor Filter Residual) features, built as histograms of quantized residuals obtained with 2D Gabor filters, can achieve competitive detection performance against adaptive JPEG steganography. In this paper, an improved version of the GFR is proposed. First, a novel histogram merging method is proposed according to the symmetries between different Gabor filters, thus making the features more compact and robust. Second, a new weighted histogram method is proposed by considering the position of the residual value in a quantization interval, making the features more sensitive to the slight changes in residual values. The experiments are given to demonstrate the effectiveness of our proposed methods.
GFR (Gabor Filter残差)特征是由二维Gabor滤波器获得的量化残差的直方图构建而成，可以与自适应JPEG隐写术相比，实现竞争性的检测性能。本文提出了GFR的改进版本。首先，根据不同Gabor滤波器之间的对称性，提出了一种新的直方图合并方法，使特征更加紧凑和鲁棒。其次，提出了一种新的加权直方图方法，考虑残差值在量化区间内的位置，使特征对残差值的微小变化更加敏感。实验证明了所提方法的有效性。
{"title":"Improving GFR Steganalysis Features by Using Gabor Symmetry and Weighted Histograms","authors":"Chao Xia, Qingxiao Guan, Xianfeng Zhao, Zhoujun Xu, Yi Ma","doi":"10.1145/3082031.3083243","DOIUrl":"https://doi.org/10.1145/3082031.3083243","url":null,"abstract":"The GFR (Gabor Filter Residual) features, built as histograms of quantized residuals obtained with 2D Gabor filters, can achieve competitive detection performance against adaptive JPEG steganography. In this paper, an improved version of the GFR is proposed. First, a novel histogram merging method is proposed according to the symmetries between different Gabor filters, thus making the features more compact and robust. Second, a new weighted histogram method is proposed by considering the position of the residual value in a quantization interval, making the features more sensitive to the slight changes in residual values. The experiments are given to demonstrate the effectiveness of our proposed methods.","PeriodicalId":431672,"journal":{"name":"Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127458853","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：