The Research On Digital Watermark Algorithms Based On Network Traffic Characteristics

With the rapid development of the Internet, network security has become more and more serious, and all kinds of attacks are emerging. Distributed Denial of Service,anonymous communication pose a huge threat in tracking the source of attack,positioning the real attack flow and network supervision. As a kind of active network traffic shaping and analysis technology, active network flow watermark gradually become the research hotspot in network security field, by virtue of its broad applicability, less time and space overhead and high detection efficiency. In recent years, according to the different watermarking carrier and traffic characteristics, the researchers put forward a number of algorithms, and achieved some results. While the existing watermarking schemes are mainly focused on carrier selection and embedding, ignoring the correlation between the watermark itself and the carrier.Besides, the same PN code spread spectrum is easy to lead to the self-similarity of traffic in multi-flow tracking. In addition, although increasing the redundancy packets can improve the robustness of the watermark to a certain extent, it also increases the time and space cost of the watermark embedding and extracting, and reduces the watermark capacity. What’s more, the continuous watermark embedding mode not only increase the probability of the watermark exposure, but also the mutual interference between the watermarks reduces the detection efficiency. Aiming at the problems existing in the watermarking scheme, based on the research on watermarking algorithm, the main research contents of this paper are as follows:First, according to the analysis and modeling of network traffic data, we complete the feature extraction of data stream by using information entropy and hash function. In this way, the flow can be identified by the characteristics of the stream itself in multi-stream tracking.Second, the idea of multiple division-frames in wireless sensor nodes is introduced into watermarking algorithm by further division the redundant packets to create multiple copies of the embedded watermark. In this section, we proposed a watermark embedding method based on the idea of double redundancy. Experimental results and theoretical analysis show that the method can effectively improve therobustness of the watermark and reduce the time and space cost in detection.Third, according to the application of discrete cosine transform in energy distribution, a watermarking embedding model based on the discrete cosine transform is proposed and verified. With the discrete cosine transform, this method can convert data flow from time domain to frequency domain, and then, we can select the embedding regions from the angle of energy. Discretely embedding model not only reduces the interference between the watermarks, but also improves the concealment and detection efficiency.In this paper, we proposed an entropy and hash based double redundancy watermark model and watermarking location method based on discrete cosine transform, on the one hand, the watermark itself is associated with the characteristics of the carrier which provides convenience for multi-flow tracing, on the other hand,through the research and improvement of watermark embedding and area selection,the robustness and concealment of the watermark is improved effectively. Moreover,the reduction of cost in detector make it possible for the application of watermark in real time environment.