15 results about "Watermark robustness" patented technology

The invention discloses an image watermark generation method and system based on a diffusion model optimization potential variable, and the method comprises the steps: training a secret encoder and a watermark extractor, and coding a message into a secret residual error; after the residual error is injected into the potential representation of the image, mapping the residual error into a potential vector with a watermark, which accords with the prior of a diffusion model, through DDIM inversion; performing iterative optimization on the potential vector by using a pre-trained diffusion model U-Net, introducing double constraints of image reconstruction loss and message fidelity loss in a denoising process, and generating an image with high visual quality and high watermark robustness; and finally, extracting the message through a watermark extractor and carrying out statistical significance verification to realize copyright authentication. According to the method, dynamic watermark embedding is achieved, the attack resistance is high, the image quality is well kept, an original model structure is not changed, and the method is suitable for copyright protection and user tracking of AIGC content.

The disclosure provides a feature point detection-based anti-screen camera robust watermark embedding, extraction method and system, relates to the technical field of digital media copyright protection, and when a carrier image is photographed and watermark is extracted, an image correction algorithm based on PCA and SURF is proposed, the generated distortion can be well corrected, and the robustness of the screen camera watermark is indirectly ensured. The image after the screen camera can be corrected to restore the image to the state in the screen, and the selection of the feature area before and after the correction is also performed to improve the embedding effect of the watermark. In order to improve the robustness of the screen camera watermark, in the embedding of the screen camera watermark, the carrier image is subjected to DCT transformation, and then experiments are designed to select the DCT coefficients of the embedded watermark. Thus, after the watermark is embedded, the robustness is high, and the invisibility is also high.

The invention provides an audio watermark embedding method and device and a conference system, and relates to the technical field of digital watermarking. And obtaining an original audio collected from an environment where the terminal is located in the first time period, wherein the original audio comprises an audio played by the terminal in the first time period. The playing audio is configured to be embedded with a watermark based on the first watermark embedding parameter value by using a first watermark embedding algorithm. And carrying out watermark extraction on the original audio according to the first watermark embedding algorithm. And if the watermark extraction from the original audio fails, embedding the watermark in the audio to be played of the terminal based on the second watermark embedding parameter value by using the first watermark embedding algorithm to obtain a new played audio. Under the condition that the watermark cannot be successfully extracted from the original audio, the watermark embedding parameter value is adjusted, so that the watermark embedded in the audio played later by the terminal can adapt to the environment and is successfully extracted, the adaptive environment adjustment of the audio watermark is realized, and the robustness of the audio watermark is improved.

This invention discloses a method and apparatus for evaluating the robustness of image watermarking. The method includes: acquiring an original image dataset; printing watermarks on a portion of the original images in the original image dataset using the watermarking algorithm to be evaluated, forming a watermarked image dataset; attacking all images in the watermarked image dataset to obtain an attacked image dataset; performing watermark recognition on all images in the attacked image dataset to obtain watermark recognition results; the watermark recognition results include the true positive rate and the false positive rate; using a preset image quality measurement algorithm, calculating the image quality measurement values ​​before and after the attack for each watermarked image, as the image quality degradation result; determining the image watermark robustness evaluation result based on the watermark recognition result and the image quality degradation result; the image watermark robustness evaluation result includes an image watermark robustness evaluation report. This invention provides a complete, systematic, and mature method for evaluating watermarking algorithms.

The invention discloses an adaptive code watermark embedding method and system based on real-time feedback, and the method comprises the steps: obtaining a currently generated code sequence in real time in a code generation process; inputting the obtained current code sequence into a pre-trained lightweight naturalness discriminator module to obtain a score reflecting the naturalness of the current code; inputting the obtained naturalness score into a dynamic bias controller module, presetting a target naturalness in the module, and calculating a watermark bias value; inputting the watermark offset value into a watermark strategy and code generation module, and generating new probability distribution after self-adaptive watermark offset adjustment; in the new probability distribution, through a lightweight naturalness discriminator, a new round of evaluation and control circulation is carried out by using the newly generated Token, a real-time feedback control loop including sensing, decision making and execution is constructed, the watermark bias strength is dynamically adjusted according to the real-time generation quality, and dynamic balance of watermark robustness and code naturalness is ensured.

The invention provides an audio watermark embedding method and device, electronic equipment and a storage medium, and belongs to the technical field of audio processing, and the method comprises the steps: obtaining the amplitude spectrum characteristics of an original audio signal, and determining a watermark embedding energy mask; inputting the amplitude spectrum features into a plurality of mask prediction network models to obtain watermark embedding risk masks; determining a target watermark embedding risk mask according to the watermark embedding risk mask to generate a frame-level embedding mask; and embedding the watermark information into the feature component corresponding to the target audio frame to obtain a watermark-containing amplitude spectrum feature, and performing inverse time-frequency transformation on the watermark-containing amplitude spectrum feature to obtain a watermark-containing audio signal. According to the method, the audio energy and the model prediction risk are combined, the audio frame which is most suitable for embedding the watermark is adaptively screened out, modification in hearing sensitive frames such as silent frames and transient frames is avoided, the problem that the audio quality is reduced due to undifferentiated embedding of the watermark is solved, and on the premise that the watermark robustness is not sacrificed, the watermark embedding efficiency is improved. And the imperceptibility of the watermark and the auditory quality of the audio containing the watermark are obviously improved.

This invention discloses a robustness enhancement method for model watermarking based on adversarial immunity, primarily addressing the problem that existing model watermarking methods are unable to resist various watermark removal attacks. The solution involves: using a public dataset suitable for image classification tasks and a convolutional neural network as the original data and target model, respectively; training the target model with the original data; using a generative model as the watermark generation model and training it with the original data; inputting the original data into the trained watermark generation model for forward propagation to output watermark data; concatenating the original data and watermark data and inputting the concatenated data into the trained target model for forward propagation to obtain the model watermark; adding minor perturbations to the original data to generate adversarial examples, concatenating these examples with the original data, and inputting them into the model watermark for robustness training, resulting in an enhanced robust model watermark. This invention effectively enhances the generalization ability of model watermarking under different data distributions, improves model robustness, and can be applied in many fields such as information security, financial services, and healthcare.

The invention discloses an image watermarking method based on human visual characteristics, which relates to the technical field of information security, and comprises the following steps: respectively converting watermark information and an original image into Lab color space, processing the original image by using an LBP operator, and layering by using texture complexity; carrying out DCT (Discrete Cosine Transform) on the three-component graph of the layer of image with higher complexity, calculating the angular frequency of the three-component graph, and calculating a contrast perception threshold value by utilizing a CSF model to obtain the maximum amount, namely the threshold value, of the embedded watermark under each frequency; performing compromise adjustment to balance the watermark embedding strength K value of each sub-block, enabling the watermark capacity to be close to a threshold value, and embedding the watermark according to the K value, the watermark information and the threshold value; carrying out inverse DCT and color space transformation on the watermarking-containing graph, and synthesizing the watermarking-containing graph with the three-component graph without the watermarking to obtain a watermarking-embedded graph; and finally, according to the K value, carrying out reverse processing on the original image and the image containing the watermark, and extracting the watermark. According to the method, the watermark transparency is better guaranteed, the watermark capacity is increased, and the watermark robustness is guaranteed.

The invention discloses a customized generative model copyright protection method based on robust watermark optimization, and the method comprises the steps: embedding a randomly generated initial watermark sequence into a basic generative model parameter, and obtaining an initial generative model with a watermark; taking the initial model as a proxy model weight, and constructing a proxy customization model to simulate an attacker customization training process; utilizing the proxy model to simulate fine tuning training on the customized data set; constructing a watermark robustness optimization target based on a gradient, and introducing a Fisher information matrix to carry out weighted constraint on a watermark updating gradient, so that a watermark parameter, namely, a robust watermark subspace, is iteratively optimized; after multiple rounds of proxy training and watermark updating are alternated, an optimized robust watermark is obtained and is re-embedded into the basic model to be published externally, and the robust watermark is used for any downstream customized generation task; in the verification stage, an image to be detected is input into the pre-training extraction network, watermark information is extracted and matched with an original watermark, and model verification and copyright evidence obtaining are completed.

The application provides an audio watermark embedding method and device, electronic equipment and storage medium, and belongs to the technical field of audio processing. The method comprises the following steps: obtaining the amplitude spectrum feature of an original audio signal and determining a watermark embedding energy mask; inputting the amplitude spectrum feature into a plurality of mask prediction network models to obtain a watermark embedding risk mask; determining a target watermark embedding risk mask according to the watermark embedding risk mask to generate a frame-level embedding mask; embedding watermark information into the feature component corresponding to the target audio frame to obtain a watermark-containing amplitude spectrum feature, and performing inverse time-frequency conversion on the watermark-containing amplitude spectrum feature to obtain a watermark-containing audio signal. The application combines audio energy and model prediction risk to adaptively select the most suitable audio frame for embedding a watermark, avoids modification in silent, transient and other auditory sensitive frames, solves the problem of audio quality degradation caused by indiscriminate watermark embedding, and significantly improves the imperceptibility of the watermark and the auditory quality of the watermark-containing audio without sacrificing the robustness of the watermark.

The invention discloses an image protection method for balancing robustness and generation diversity. The method comprises the following steps: encoding a randomly generated random session key by adopting a preset master key to obtain a random session key code; encoding a watermark to be embedded for image protection by using the random session key to obtain a watermark code; and generating a watermark noise vector based on the random session code and the watermark code, and embedding the watermark noise vector into a to-be-protected image. According to the method, a robust and hidden generation image detection and traceability scheme is realized, so that the watermark embedded in the image can give consideration to the robustness of the watermark and the diversity of image generation.

The invention provides an anti-screen shooting image watermark network method and system based on double-domain fusion and frequency domain attention. The method comprises the following steps: step 1, preprocessing binary watermark information; step 2, constructing a double-domain collaborative high-frequency residual error enhancement network DRHE-Net; 3, constructing a real noise-containing watermark image for network training; step 4, establishing a frequency domain attention decoding network matched with the encoder; and step 5, a weighted multi-objective loss function is adopted to complete network end-to-end training, and balance between imperceptibility and robustness of watermarks is realized. According to the method, the watermark information and the carrier image are processed by using the double-domain collaborative residual enhancement network and the frequency domain attention mechanism, and accurate balance between watermark robustness and imperceptibility in the screen shooting scene is realized.

The invention discloses a large language model watermark embedding and detecting method based on attention head perception, and relates to the technical field of natural language processing and digital watermarking. The method comprises the following four core steps of: firstly, screening an attention head set which plays a key role in factual information coding through a'pure-pollution-recovery 'three-stage causal intervention experiment; secondly, a binary linear fact detector is trained based on the set, and accurate distinguishing of token factuality and non-factuality is achieved; then, in a model reasoning stage, a dynamic green list is only generated for non-factual tokens, watermark hidden embedding is completed through logits bias intervention, and bias intensity is adaptively adjusted along with text fact density; and finally, in a detection stage, reproducing an embedding process to screen non-factual tokens, and quantifying an abnormal degree by counting a matching ratio and a z score to realize accurate detection of the watermark. The text factual accuracy is prevented from being damaged through the attention head perception mechanism, the watermark robustness and concealment are improved through the dynamic green list design, the method is suitable for scenes such as copyright tracing and authenticity verification of the text generated by the large language model, and the engineering feasibility is high.

The invention discloses an image processing method and device, equipment, a storage medium and a program product, and is applied to the technical field of image processing. The method comprises the following steps: acquiring a first image and digital watermark content; the first image is input into a watermark calibration model, an image content confidence degree mask image of the first image output by the watermark calibration model is obtained through the watermark calibration model, and the image content confidence degree mask image is an image used for representing the attention degree of the watermark calibration model to different areas in the first image; according to the image content confidence mask image, the position where the digital watermark content is embedded into the first image is determined, and the value of the attention degree corresponding to the first position in the image content confidence mask image is larger than or equal to the value of the preset attention degree; and based on the first position, performing fusion on the digital watermark content embedding and the first image to obtain a second image after the digital watermark content is embedded into the first image. In this way, the watermark robustness is improved by dynamically selecting the high-robustness embedding position.

The invention discloses a digital sound watermark processing method and device based on echo hiding and electronic equipment, and relates to the technical field of digital watermarking. According to the core scheme of the digital sound watermark processing method based on echo hiding, a first audio signal is segmented through duration, then the energy of each segment is calculated, watermarks are embedded into the segments with high energy through an echo hiding method, and watermarks are not embedded into the segments with low energy. In this way, the shielding effect of human auditory sense is utilized, that is, the strong signal can cover the adjacent weak signal, so that the imperceptibility after the watermark is embedded is improved while the robustness of the watermark is ensured, that is, the relatively high audio subjective quality is kept. Meanwhile, an effective way is provided for controlling the embedding capacity of the watermark.