Research On Audio Fingerprinting Algorithm Of Anti-collusion Attack Based On Frequency Domain Flip

The rapid development of computers has brought unprecedented convenience to the sharing and dissemination of multimedia data.It has also brought great risks and hidden dangers.For example,the dissemination of unauthorized works not only brings copyright infringements and property losses to the copyright owners of the works,but also hinders the development of the multimedia copyright industry.Copyright protection is particularly important in the current situation of convenient and rapid data dissemination.Information hiding technology plays a very important role in digital signature,copyright protection,tamper detection,etc.It is proved by embedding some identifying information in multimedia data and extracting it when necessary.As a very important application branch in the field of information hiding,the development of digital fingerprint technology has become a hot topic for many scholars.The data fingerprint technology embeds different fingerprints data information in the works and then distributes them to users who have obtained the usage rights through purchase.Therefore,the final copy obtained by each user is unique due to the difference in fingerprint data.Digital fingerprint technology can effectively solve the problem of some users colluding with works driven by interests to further damage the copyright of works.However,most of the existing schemes design fingerprint codes according to certain rules.When collusion occurs,fingerprints can be traced to the users participating in the collusion based on the works produced by the collusion.In this paper,based on the in-depth study of data fingerprinting technology and the principle of collusion attack and the result of the damage caused,an effective solution to the collusion attack is proposed by fliping the frequency domain coefficient of the audio frequency with a specific step length,which does not need to track the illegal users and fundamentally solves the problem of collusion attack.In order to further improve the robustness of the algorithm,a scheme combining frequency domain flipping and histogram modification is proposed,which combines the processing methods of time domain and frequency domain.The research on audio copyright protection against collusion mainly includes the following two tasks:(1)A fingerprint embedding algorithm based on frequency domain over-flip is proposed.Due to the attack performance of the collusion attack,the embedded fingerprint is changed to eliminate the fingerprint trace,and a targeted embedding method is proposed.Fourier transform is used to perform frequency domain transform processing on audio signals,and the frequency domain coefficient segments are flipped to different degrees according to the fingerprint sequence.In the process of flipping,in order to improve the performance of the scheme and increase the discrimination of different fingerprint copies,a misalignment value is set during segmentation.Complete the embedding of audio fingerprint information under conditions that are not perceivable,and prevent the duplicates produced by dishonest users from colluding from being published again.This solution not only achieves the difference between each copy,but also makes the content of the frequency domain different.When users collude,the frequency domain modified by the embedded information in the copy becomes relative,and the mutual offset makes this method from the root cause Elimination of collusion,effectively avoiding the copyright hazards caused by the second release,is of great significance to the copyright protection of audio.When multiple authorized users want to form new copies based on various copy collusion attacks for the illegality release of works,they will find that the quality of the collusion version is severely damaged and cannot be used for secondary use.(2)An embedding scheme based on frequency domain flipping and histogram modification is proposed.The scheme includes two layers of embedding mechanism.The first part is based on frequency domain inversion processing,in which the fingerprint can resist collusion attack;the second part is based on the modification of square graph.The purpose of this step is to further improve the robustness of the scheme.The histogram is modified to embed the information.Every three adjacent Bins are divided into a group,and the data relationship between bins is modified according to the embedded bit information to meet the embedding conditions.Histogram based modification is not related to the location of the sample,only statistical data features,robust performance is better.Therefore,this scheme combines the advantages of the two embedding mechanisms to further improve the performance of the scheme.The scheme can realize a good balance between imperceptibility and robustness.