Achieving Sustainable 5GView this Special Issue
Research Article | Open Access
General Multimedia Trust Authentication Framework for 5G Networks
Due to the varieties of services and the openness of network architectures, great challenges for information security of the 5G systems are posed. Although there exist various and heterogeneous security communication mechanisms, it is imperative to develop a more general and more ubiquitous authentication method for data security. In this paper, we propose for the 5G networks a novel multimedia authentication framework, which is based upon the trusted content representation (TCR). The framework is general and suitable for various multimedia contents, e.g., text, audio, and video. The generality of the framework is achieved by the TCR technique, which authenticates the contents’ semantics in both high and low levels. Analysis shows that the authentication framework is able to authenticate multimedia contents effectively in terms of active and passive authenticating ways.
The requirements for higher data transmission rate and better user services experiences promote the development of the 5G cellular networks. Compared with its precedent wireless communication standards (i.e., 1G, 2G, 3G, and 4G networks), the 5G system is proposed to support wider range of connected devices types and various service applications [1, 2]. It is estimated that the 5G would become the ubiquitous information infrastructure network in the near future . Along with the growth of the users and services in 5G networks come the security problems, e.g., data privacy, data integrity, and data authentication [4–6]. Ensuring the secure wireless communication, together with the security of data transferred, is mandatory for the success of the 5G networks.
Various types of multimedia content proliferate over the wireless networks. Especially equipped with the social media techniques, people find it is quite convenient to communicate on the mobile devices. Although the multimedia contents bring much convenience for the information sharing over the 5G networks, security concerns of the contents cannot be overlooked [7–9]. Due to openness feature of the underlying network, the received multimedia contents should be carefully examined. For example, questions should be raised whether the obtained video or image has been attacked by malicious purposes, or whether the video has been deliberately altered to screw its original meaning. Thus the security of multimedia contents must be ensured to keep the integrity of the contents as intact as the original.
As far as the formats of the multimedia are concerned, methods to authenticate the contents are different in terms of purpose, efficiency, and validity. Those methods or algorithms for authenticating the multimedia contents can be roughly classified into four groups, i.e., watermarking-based, encryption-based, streaming-based, and robust-hashing-based methods.
The watermarking-based authentication methods for multimedia usually adopt the watermark embedding and extraction approaches. For example, Md. Asikuzzaman et al.  proposed an imperceptible and robust blind watermarking for video, in which the watermark was embedded into the levels of the dual-tree complex wavelet transform coefficients. There are other multimedia watermarking methods which emphasize the features of the watermark, e.g., the content dependent video watermark  and wavelet-based multiple watermarks . Although the watermarking methods can provide the authentication of contents in some degree, the embedding process itself harms the integrity of the content. In addition, one watermarking method can only deal with some specific attacks, which hinders these methods from being widely adopted to cope with the various attacks existing in the 5G networks.
The encryption-based multimedia authentication methods usually employ the encryption algorithms, e.g., RSA algorithm and DES algorithm, to encrypt the whole or parts of the contents. For example, Zafar Shahid et al.  presented a selective encryption (i.e., the truncated rice code and the Exp-Golomb code) approach for video of High Efficiency Video Coding (HEVC), which shows the characteristics of format compliant and real-time property. Similarly, Glenn Van Wallendael et al.  proposed the encryption for HEVC by utilizing the differences of intraprediction mode, the sign of motion vector difference, and the residual sign. Note that the primary goal of the encryption-based methods is to ensure the confidentiality. The time overhead incurred by the encryption and decryption process is often a nonnegligible issue for the social media devices.
The aim of the streaming-based multimedia authentication methods is to ensure that the received contents have the integrity property, which is verified by the recipients. Currently many research works have been focused on this area; e.g., Kang et al.  studied the pollution attack detection and prevention method by trust for peer-to-peer streaming, Lu et al.  proposed a privacy protection method based on trust for peer-to-peer data sharing network, and Cheng et al.  authenticated the live streaming media by means of TESLA- (Timed Efficient Stream Loss-tolerant Authentication-) based protocol. One category of streaming-based authentication methods is the graph structure type, e.g., the chain line approach, the tree approach, and the butterfly approach, which treats the content packets as individual ones and exploits various packets’ hash attaching means . The other category is based on the multimedia coding structure. For example, Kianoosh Mokhtarian et al.  proposed the authentication for Scalable Video Coding (SVC) streams, whose hash appendence follows the coding structure of SVC.
The robust-hashing-based methods to authenticate the multimedia have the best performance in terms of robustness when the authentication is applied under the network circumstances. The robustness of the hashes means that hashes remain the same or change a little after the perceptual content preserving operations are being conducted. For example, Lokanadham Naidu Vadlamudi et al.  proposed a robust hash algorithm by using features of histogram for image authentication. Due to the robustness and sensitivity of this kind of method, it is much superior to the other three methods when it comes to the network packet loss phenomenon in 5G networks. Since certain packets of multimedia contents may be lost because of traffic jams or network failure, contents integrity verifying process for the recipients should take into consideration the robustness of multimedia’s representation.
Another aspect we should bear in mind is that all the four methods are for the low level semantics authentication, which treats only the integrity either in pixels level or in frames level. This absolutely lacks completeness for the authentication of multimedia contents of the 5G networks, since high level semantics of the contents are overlooked by the current methods. For instance, the contents’ high level semantics including but not limited to title, name, author, and format should also be authenticated. Therefore, in this paper we endeavor to tackle the multimedia authentication by proposing combined authentication method for multimedia contents for 5G network. We propose a generalized multimedia trust authentication model based on the trusted content representation (TCR) method, where the generalized term means that it can be applied to various multimedia content formats. Also we provide the features needed to authenticate the integrity of the contents.
The rest of the paper is organized as follows. We describe the architecture of asymmetric wireless communication channel in Section 2, where the asymmetric term means that one channel is safe and is used for authentication information transfer while the other one is open and used for multimedia contents transfer. In Section 3 we illustrate the TCR indexing technique and show the features of TCR. In Section 4 we explain the framework of the general multimedia trust authentication for 5G networks, in which the philosophy of asymmetric wireless channel is adopted. Then we conclude our work in Section 5.
2. Architecture of Asymmetric Wireless Communication Channel
2.1. Security Threats to 5G Networks
Since the 5G networks are open and insecure, the multimedia contents, which are transmitted over them, are prone to various attacks by malicious purposes. From the perspectives of network communication system, we summarize security threats to multimedia into three groups.
(1) The Originality of the Multimedia May Lack Trust. Currently users can upload from the mobile devices the various kinds of multimedia contents up to the video or audio sharing websites by various means. However, the effective and efficient mechanisms to supervise the legacy of users’ behaviors are still unavailable. While people find it quite easy to collect, edit, and distribute the multimedia contents by modern multimedia processing tools (e.g., Photoshop, Illustrator), people should think twice about the originality of the received multimedia. In other words, the user, who publishes the multimedia and claims the author of the multimedia, may likely not be the “true” author. Originality of the contents should be checked in some manner. This incurs the problem of intellectual property infringement. Therefore, it is urgent to find a good way to verify the originality of the multimedia.
(2) The Channel to Convey the Multimedia May Be Attacked. The openness of the 5G network protocols poses serious security problems to the contents being transmitted over it. The notorious attack, “man-in-the-middle”, is the most common threat to the contents. The attacker maliciously hurts the multimedia via many means, e.g., video frames’ rearrangement, dropping, inserting, and altering. The purpose of the attack is to destroy the contents’ integrity, which distorts the recipient’s understanding of the contents. Note that there are many free packets sniffer tools on the Internet, which makes the attacks on the transmission process of multimedia much easier. We state that the threats to the multimedia on its journey from the sender to the receiver should be taken into consideration and the integrity of the contents needs to be ensured.
(3) The Identity of the Multimedia Recipients May Lack Trust. Recall that current authentication for the mobile recipient’s identity is mainly based on the user-password method. This brings security problems once the legal recipient’s platform is attacked. For instance, an attacker may break into and obtain the legal recipient’s password to certain multimedia server, which results in the illegal copy or broadcasting of the contents. Another scenario is that the multimedia sender requires checking the identity of the receiver and the integrity of contents received by the receiver, e.g., video conferencing. Therefore, both the platform security and the received multimedia integrity are necessary to be authenticated.
2.2. The Asymmetric Wireless Communication Channel
The concept of asymmetric wireless communication channel for 5G networks is shown in Figure 1. We use the term asymmetric to describe the unbalanced properties of two channels existing in the architecture. One channel is the ordinary wireless communication channel, which is reserved for the multimedia contents sharing. The channel is dual and its capacity is large enough to support various kinds of services. It is not transparent to the users and not attacks-free due to its openness characteristics. The other channel, which is more akin to logical broadcasting communication channel, is introduced to transfer the authentication information used for multimedia contents. This channel is single-way and transparent. It does not exist in current networks and should be created by means of broadcasting techniques. Its capacity is much less compared to the former one. However it is closed and safe since the information transmitted over this channel is confidential.
The Trusted Content Authentication center (TCA center) in Figure 1 serves as a trusted third party. It collects the multimedia content information from the 5G network base stations. For instance, it analyzes the history records of users’ behaviors and extracts multimedia content information. Then it generates the trusted content representation for each content resource. The representations of authentication are the safety benchmarks for the multimedia contents. We apply the logical broadcasting channel to safeguard these representations. The term logical means that it can be realized by virtual technologies, such as the network tunnels.
Suppose the representations are safely transferred to the User Required Authentication center (URA center), then the remaining task is to authentication the multimedia contents according to users’ authentication requirements. The terminal is the wireless devices and is responsible for communicating with the center. It can have an application installed within its platform and initiate the authentication request to the URA center. The center returns the authentication information, together with the multimedia content profile, to the terminal, which are wrapped as information aggregation. If the center is not capable of providing relevant authentication information, it sends to the terminal a replay indicating that the terminal asks the TCA center through the wireless communication channel for the authentication information. The TCA center forwards the corresponding authentication information to the UCA center, which further sends it to the terminal. Then the terminal checks the multimedia contents’ integrity by comparing the representation with the content itself.
According to the proposed asymmetric channels for the 5G networks, there are two ways for the users or terminals to obtain and authenticate the multimedia content information. The terminal connects to the wireless base station, which provides the access point to the Internet. The users upload or download contents through these points. In order to ensure the security of the terminal, the access points apply authenticity test to the terminals. The TCA center forms the TCR index for the multimedia content. Considering the long-tail features of the popularity distribution of the content, we only generate the indexes for those “popular” contents, which can satisfy almost all users’ authentication requirements [21, 22].
The other channel for the user to obtain the authentication information is the logical broadcasting band. The application on the terminals asks the URA center for contents authentication information by specifying the contents profile. Then the center searches its authentication information database for the specified terms. If it finds the required one, then it returns back to the terminal. Note that in order to protect the security of wireless channel between the terminal and the center, encryption methods are to be adopted. We have researched the asymmetric channel in our previous work and its detail is referred to [23, 24]. The database, which stores the authentication benchmarks, receives the authentication information from the TCA center periodically. The security of this logical channel must be ensured in order to provide the basis of trust for terminal authentication test.
3. Multimedia Trusted Content Representation
3.1. Trusted Content Representation Description
We first clarify what information the recipients can obtain from the multimedia contents for 5G networks. As for the contents recipients, they can get two kinds of information, i.e., the high level semantics and the low semantics. The former means that the descriptions about the contents, while the latter means the perceptual understanding of the contents. We define these two terms in detail as follows.
(1) Multimedia High Level Semantics (HLS). This kind of semantics is the description or explanation information about the multimedia, which is usually generated by the content author and is mainly used for the indexing and searching. The semantics are in the form of plaintext and can be added before the contents. Examples of HLS are title, author, keywords, and date. Note that HLS belongs to the conceptual level regarding recipients understanding of the contents.
(2) Multimedia Low Level Semantics (LLS). The LLS denotes the multimedia data which provides perceptual information for the recipients. They often exist in the form of binary data, e.g., the pixel matrix of an image and the frame series of a video. Note that information of LLS needs no summary of conceptual levels by contents authors or publishers. The perceptual understandings are formed by the contents recipients themselves.
We believe that the authentication of multimedia contents should cover both HLS and LLS information. Specifically, the HLS of contents should remain the same as the one of original contents and the LLS should also be as intact as the original one. The undergoing operations on the LLS, which are without malicious purposes, should preserve the perceptual understanding as the original one. Considering the 5G network packets loss to the data of LLS, we propose to authenticate the LLS robustly. We believe the robust-hashing is more appropriate for 5G networks and choose the robust-hashing methods to generate the LLS of contents.
Note that, as for the HLS, we suggest no robustness to its description. One reason is that the HLS information is quite sensitive and one bit change may completely destroy its meaning. The other reason is that we use the HLS to locate the multimedia contents and the HLS serves as the authentication starting points for contents verification. Therefore, we propose the TCR method, which represents the HLS and LLS robustly. The model is described as follows.
We denote the multimedia contents space by M and the multimedia content by m. Let the HLS of M be denoted by S and we separate the space S into several subspaces, which are mutually independent of each other; i.e., the space S is the cross product of its subspaces. Each subspace represents certain conceptual description of M. Then the HLS for m is depicted by s, which is defined asThe terms of s are the instances of the subspaces of S. Note that the number n is an integer which represents the granularity of the HLS spaces.
Regarding the robustness representation of the LLS for multimedia contents, we allow the perception preserving operations on the contents on condition that these operations do not change the representation too much. However, the degree of change is determined by the authentication requirements relating to the network circumstances. In order to clearly characterize the LLS feature, we here examine the categories of operational results of content m.
We adopt the robust-hashing authentication method for the LLS representation. For the content m, the results of operations being conducted on m can be thought of having three groups, i.e., , , , which are explained as follows.
(1) The set denotes the multimedia contents, which are produced by the perception preserving operations on the content m, e.g., the scaling of image and the brightness change on the video frame.
(2) The set denotes the multimedia contents, which are produced by the content changing operations on the content m, e.g., the frames reordering of video and the image blocks covering.
(3) The set denotes the multimedia contents, which are different from and independent of content m. The LLS of is completely and fundamentally distinguished from m.
We apply the robust-hashing methods to all those above sets. Let the hashing function be denoted by H(·). Thus we obtain the corresponding hashes as follows.Obviously we have the relation , where H denotes the LLS space for multimedia contents.
We propose the trusted content representation method to describe the HLS and LLS for content m. Therefore the content m is represented by T, which is defined asThe TCR model is depicted in Figure 2, in which the HLS space is composed of S and . The is the HLS for contents which is different from and independent of content m. Note that, as for the LLS, we allow the robustness, which means that the hash of the content being verified against m is secure if it falls into the set . The parameter ε decides the distance between the original hash and the allowed maximum hash, which also draws the line for the robustness of the LLS for content m.
In regard to the authentication for multimedia content m, we apply the TCR technique to verify the received content. We first check the extracted HLS to see if it equals S. If no, we conclude that the received content is not safe since the high semantics are not secure; if yes, we further check the extracted LLS to see whether it is within the range of . If yes, we state that the content being verified is authenticated; if no, we assert that the low level semantics are being attacked and the content received is not secure.
3.2. Properties of Trusted Content Representation
We in this part summarize the properties of the model when it is employed in the 5G networks to authenticate multimedia contents. There are five properties, i.e., collision-free, security, robustness, sensitivity, and compactness.
(1) Collision-Free Property. The collision-free property means that, for every two dissimilar multimedia contents, they have different trusted content representations. Suppose that and , and contents m and are different, then we havewhere symbol means logic operation AND and means the probability. The parameter is determined theoretically or empirically and it defines the robustness of the LLS for the contents. In other words, the distance for every two different multimedia contents is greater than the threshold with probability one.
(2) Security Property. This property ensures that the trusted content representation for multimedia contents should be resilient against various malicious attacks and be detectable after being attacked. It also states that the representation T for certain content cannot be regenerated by attackers. Usually this is achieved through key controlled representation generation. Specifically, the LLS should be hashed by keys. The security property has two aspects; i.e., one is the one-way hash, and the other is the unpredictability.
The one-way hash means that it is easy to generate the hash for LLS from the contents, while it is extremely difficult and impossible to recover the LLS of contents from the hash value. Suppose the contents m has the representation (S H), and from T can be inferred the content m’, then we have where the integer L is the length of the hash and we assume the one and zero in the hash follow uniform distribution.
The unpredictability means that given a multimedia content m, it is empirically impossible to infer its hash for LLS. Since the hash for LLS generation is controlled by key, suppose key is used for the legal generating LLS hash for m and key is another key which is illegal, then we havewhere means that the hash generation is controlled by K1. Note that K1 does not equal K2.
(3) Robustness Property. The robustness property of this model is embodied in the LLS for the contents; i.e., some conception preserving operations on the LLS are permitted. Suppose we have a content m and its similar version , then we have the following relationwhere threshold draws the boundary between similarity and difference for contents comparison.
(4) Sensitivity Property. This one is quite important for the contents security verification. It means that the model can detect those attacks which result in the changes of trusted content representation. There are two kinds of attacks regarding the semantics of multimedia. One is attacks on the HLS and the other is on the LLS. The latter attacks harm recipients’ perceptual understanding on the LLS. We denote the altered contents of m by , which belongs to the set . Then the following relation holds.where the symbol denotes OR logic operation.
(5) Compactness Property. Compactness property means that the shorter the trusted content representation is, the better it is. If the length of the representation is shorter, it requires less storage and less transmission overhead for 5G networks. However, we believe that there is a balance between the compactness and correctness of the representation. For instance, if the contents’ HLS representation S has much larger length, then S can represent more detailed information. It means the HLS space is divided into subspaces of more granularities. On the contrary, if the length is smaller, the representation S is much more compact. But more HLS information cannot be included. Therefore, it needs tactics and empirics to decide the length of the trusted content representation. Note that this balance also applies to the LLS representation.
Although these five properties exist in the TCR model, we claim that not all these five properties can be met at one time. For example, if the model is more emphasized on the security of the multimedia contents, then the security and sensitivity properties should come first. However, if the model is more concerned from the perspective of the transmission efficiency, then the compactness property should be weighed more. Thus in real-life applications of 5G system, the properties of the model should be determined accordingly.
4. General Multimedia Trust Authentication Framework
When we apply the TCR model in the 5G networks, we break this authentication into two processes, i.e., the semantics analysis and adding by the sender, the semantics extraction and comparison by the recipient, which is shown in Figure 3.
We denote the raw multimedia content generated by sender by . First we analyze the content to obtain its HLS and LLS semantics, which are represented by symbols S and H, respectively. The HLS is added to through semantics management. The adding method can be realized by the appendence to the content’s head or tail. The multimedia , together with S, is conveyed to the recipient across the channel. The channel is an open network, i.e., the 5G network, whose security is not ensured. On the other hand, the HLS and LLS are synthesized into trusted content representation T, which is securely transmitted by safe channel. The safe channel can be achieved by PKI (Public Key Infrastructure) technique, which encrypts the representation T through the certificates provided by an authentication center.
The safe channel is also used to test the recipient’s identity. Note that the safe channel serves as the logical broadcasting channel depicted in Figure 1. Before the recipient can access any contents from the 5G networks, its authenticity must be ensured by the base stations. Some modern identity method, e.g., trust computing, can be applied to check the security profile of the terminals’ platform. Thus the base station first initiates the identity authentication requirement when the recipient asks for connecting to the wireless networks. It starts the challenges and responses mechanism to perform the credit checking. For instance, the terminal hardware stores certain credit issued by the base station parties. The base station sends the nonce to the terminal. Then the terminal returns the cipher message of credit and nonce, digitally encrypted by the station’s public key, to the station. The base station decrypts the cipher to compare the recipient’s credit with its stored one and decides whether the recipient is legal or not. It also compares the decrypted nonce to its previously sent one. If they match, then it concludes that the message is new and free of replay attack. Otherwise, it believes the message is generated by attackers and rejects the access requirement. We use the base station to actively verify the authenticity of the recipients in order to prevent some malicious users from deliberately sabotaging the security of the 5G networks.
The recipient obtains from the channel the content denoted by M’, whose integrity is to be verified by the recipient. The recipient extracts from M’ the HLS and LLS semantics and form the trusted content representation T’. Then the recipient compares the representations T’ and T and judges whether the content M’ is secure or not. If those two terms are the same, the recipient can be ensured that the content M’ is of integrity. Otherwise, it is not secure and it may have been attacked. Note that when the recipient compares the LLS, the robustness should be taken into consideration.
Regarding the semantics management, we propose a metadata based method to append the HLS to the multimedia content. The metadata is organized in XML format. It presents its semantics to the recipient before the content is presented to the recipient. For example, the packets of supplemental enhancement information within the H.264 stream can be used to store the HLS; the label unit of “APPn” within the JPEG file can be used for the HLS. We provided a method to divide the HLS space into 14 subspaces based on our previous work . The instance of XML based HLS representation is shown in Figure 4, in which fourteen concepts are used for characterizing the high semantics.
We observe from the framework of Figure 3 that there are both the active and passive authentication modes in our model. The HLS information authentication is active, since it adds or appends additional semantics information into the multimedia contents. However, the LLS authentication is passive, because we extract the content’s low level conceptual description akin to the digest. The superiority of our framework is that it does not harm the content itself and the trusted content representation can be transferred separately from the content, which can be protected by advanced encryption algorithms.
5. Security Analysis of the Framework
Suppose that recipient obtains from the 5G networks the multimedia content lacking trust. This could be caused by attackers modifying the semantics of the content. We denote the “original” content uploaded by an attacker to the networks by TCR T’, which generates the HLS and LLS as and , respectively. The real and secure content relating to the T’ is denoted by T, which consists of HLS and LLS, respectively. The purpose of the attacking is to distort the semantics of the content. It can be realized from three perspectives. The first one is that the attacker modifies the high semantics of the content while keeping the low semantics intact. The second case is opposite compared to the first, attacking the low semantics of the content while keeping the high semantics intact. The third is that attacker modifies both the high and low semantics.
Note that in our proposed trust authentication framework the safe channel is to transfer the authentication information. Thus we assume this channel is free of attacks and the destination of the channel can obtain the secure and intact information. We analyze the security of the framework from three aspects according to the types of attacks to multimedia of the 5G networks.
For the first scenario, the high semantics is attacked only. Thus we have and . The terminal obtains the TCR from the URA center in Figure 1 as (). The H could be valid or void, because there is a chance that no content exists under the high semantics . If H is void, the terminal can claim that the content it received is not trusted. If H is valid, then it should have according to the collision-free property. Thus terminal also can find the attacks. Therefore, the authentication is successful under this kind of attack scenario.
For the second type attack, only the low semantics is altered. Thus it holds that and . This can be easily be detected since the terminal can obtain the TCR for the received content as (). According to the robustness and sensitivity properties, the terminal can be alarmed that . Note that the aim of this attack is to distort the low semantics and thus the difference between and H is large enough to be detected. Otherwise, the falls into the robustness property of H and the low semantics of the attacked version could be almost the same as the original one. This attack scenario can also be successfully detected.
For the third kind of attack, both the high and low semantics are changed. We have and . It means that the attacked multimedia content is completely different from the original one. By comparing the HLS of the content, the URA center could return TCR () to the terminal if by chance the attacked content is the same as other legal contents. But this chance could be almost impossible since we suppose the attack is to distort the semantics and not to change the content into another legal one. Similar to the first attack scenario, H could be void. Thus the terminal can know that the content lacks trust.
The above analysis shows that the general multimedia trust authentication framework can detect the attacks which have already happened. It authenticates the contents in a passive way. However, it can also authentication the content actively, which means that it can prevent the attacks from happening. Note that the base station first verifies the credit of the terminal before it can access the resources of the 5G networks. If certain attacker pretends legal user and asks for permission to the network, it lacks the credit and will fail to be part of the wireless network. However, we state that our framework cannot stop the legal terminal from harming the 5G network security. For instance, the legal user could download the legal content, deliberately alter the content, and reupload the content. However, we believe the legal user or terminal could behave securely and legally and this scenario is beyond our analysis for the authentication framework.
The secure and sustainable architecture of the 5G networks is vital for the networks health developments. We propose a general multimedia content trust authentication framework to verify the various categories of multimedia. The framework is novel in that it adopts two channels to transmit the information; i.e., one is for the usual multimedia and the other is for the authentication information. The former one is open and prone to attacks while the latter is closed and free of attacks. The framework adopts the trusted content representation technique, which authenticates the contents in both high and low level semantics. The high level semantics are conceptual terms generated to locate the contents from humans’ understanding. The low level semantics are robust and perceptual terms to measure the integrity of perception. We analyze the security of the framework and show that it can authenticate the multimedia contents actively and passively. In our future work, we will look at the implementation of the general multimedia contents trust authentication by simulating upon 5G systems. Attention should also be focused on the attacks and security experiments on the proposed model.
The data used to support the findings of this study are available from the corresponding author upon request.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
This research is supported by National Natural Science Foundation of China (Grant no. 61771185, Grant no. 61772175), Doctoral Research Foundation of Southwest University of Science and Technology (Grant no. 17zx7158), Science and Technology Research Project of Henan Province (Grant no. 182102210044, Grant no. 182102210708), and Key Scientific Research Program of Henan Higher Education (Grant no. 18A510009, Grant no. 17A520005).
- A. Tzanakaki, M. Anastasopoulos, I. Berberana et al., “Wireless-optical network convergence: Enabling the 5G architecture to support operational and end-user services,” IEEE Communications Magazine, vol. 55, no. 10, pp. 184–192, 2017.
- P. K. Agyapong, M. Iwamura, D. Staehle, W. Kiess, and A. Benjebbour, “Design considerations for a 5G network architecture,” IEEE Communications Magazine, vol. 52, no. 11, pp. 65–75, 2014.
- L. Yan, X. Fang, and Y. Fang, “A Novel Network Architecture for C/U-Plane Staggered Handover in 5G Decoupled Heterogeneous Railway Wireless Systems,” IEEE Transactions on Intelligent Transportation Systems, vol. 18, no. 12, pp. 3350–3362, 2017.
- N. Yang, L. Wang, G. Geraci, M. Elkashlan, J. Yuan, and M. Di Renzo, “Safeguarding 5G wireless communication networks using physical layer security,” IEEE Communications Magazine, vol. 53, no. 4, pp. 20–27, 2015.
- R. Chaudhary, N. Kumar, and S. Zeadally, “Network Service Chaining in Fog and Cloud Computing for the 5G Environment: Data Management and Security Challenges,” IEEE Communications Magazine, vol. 55, no. 11, pp. 114–122, 2017.
- P. Gandotra and R. K. Jha, “A survey on green communication and security challenges in 5G wireless communication networks,” Journal of Network and Computer Applications, vol. 96, pp. 39–61, 2017.
- L. Xing, Q. Ma, and L. Jiang, “Microblog user recommendation based on particle swarm optimization,” China Communications, vol. 14, no. 5, pp. 134–144, 2017.
- Q. Ma, L. Xing, and L. Zheng, “Authentication of Scalable Video Coding Streams Based on Topological Sort on Decoding Dependency Graph,” IEEE Access, vol. 5, pp. 16847–16857, 2017.
- L. Xing, Z. Zhang, H. Lin, and F. Gao, “Content Centric Network with Label Aided User Modeling and Cellular Partition,” IEEE Access, vol. 5, pp. 12576–12583, 2017.
- M. Asikuzzaman, M. J. Alam, A. J. Lambert, and M. R. Pickering, “Imperceptible and robust blind video watermarking using chrominance embedding: a set of approaches in the DT CWT domain,” IEEE Transactions on Information Forensics and Security, vol. 9, no. 9, pp. 1502–1517, 2014.
- I. Setyawan and I. K. Timotius, “Content-dependent spatio-Temporal video watermarking using 3-dimensional discrete cosine transform,” in Proceedings of the 2013 5th International Conference on Information Technology and Electrical Engineering, ICITEE 2013, pp. 79–83, Yogyakarta, October 2013.
- B. Sridhar and C. Arun, “An enhanced approach in video watermarking with multiple watermarks using wavelet,” Journal of Communications Technology and Electronics, vol. 61, no. 2, pp. 165–175, 2016.
- Z. Shahid and W. Puech, “Visual protection of HEVC video by selective encryption of CABAC binstrings,” IEEE Transactions on Multimedia, vol. 16, no. 1, pp. 24–36, 2014.
- G. Van Wallendael, A. Boho, J. De Cock, A. Munteanu, and R. Van De Walle, “Encryption for high efficiency video coding with video adaptation capabilities,” in Proceedings of the 2013 IEEE International Conference on Consumer Electronics, ICCE 2013, pp. 31-32, Las Vegas, USA, January 2013.
- X. Kang and Y. Wu, “A trust-based pollution attack prevention scheme in peer-to-peer streaming networks,” Computer Networks, vol. 72, pp. 62–73, 2014.
- Y. Lu, W. Wang, B. Bhargava, and D. Xu, “Trust-based privacy preservation for peer-to-peer data sharing,” IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol. 36, no. 3, pp. 498–502, 2006.
- C. Cheng, T. Jiang, and Q. Zhang, “TESLA-based homomorphic MAC for authentication in P2P system for live streaming with network coding,” IEEE Journal on Selected Areas in Communications, vol. 31, pp. 291–298, 2013.
- A. Habib, D. Xu, M. Atallah, B. Bhargava, and J. Chuang, “A tree-based forward digest protocol to verify data integrity in distributed media streaming,” IEEE Transactions on Knowledge and Data Engineering, vol. 17, no. 7, pp. 1010–1013, 2005.
- K. Mokhtarian and M. Hefeeda, “Authentication of scalable video streams with low communication overhead,” IEEE Transactions on Multimedia, vol. 12, no. 7, pp. 730–742, 2010.
- L. N. Vadlamudi, R. P. V. Vaddella, and V. Devara, “Robust hash generation technique for content-based image authentication using histogram,” Multimedia Tools and Applications, vol. 75, no. 11, pp. 6585–6604, 2016.
- Y.-J. Park, “The adaptive clustering method for the long tail problem of recommender systems,” IEEE Transactions on Knowledge and Data Engineering, vol. 25, no. 8, pp. 1904–1915, 2013.
- M. O'Mahony and M. Bray, “The long tail of content,” Communications Engineer, vol. 4, no. 4, pp. 20–25, 2006.
- L. Xing, J. Ma, X.-H. Sun, and Y. Li, “Dual-mode transmission networks for DTV,” IEEE Transactions on Consumer Electronics, vol. 54, no. 2, pp. 474–480, 2008.
- L. Xing, L. Jiang, G. Yang, and B. Wen, “A novel trusted computing model for network security authentication,” Journal of Networks, vol. 9, no. 2, pp. 339–343, 2014.
- L. Xing, Q. Ma, and M. Zhu, “Tensor semantic model for an audio classification system,” Science China Information Sciences, vol. 56, no. 6, pp. 1–9, 2013.
Copyright © 2018 Ling Xing et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.