Short video active defense encryption system based on device fingerprint and dynamic confusion field

Abstract

The application relates to the technical field of short video encryption, and discloses a short video active defense encryption system based on device fingerprints and dynamic confusion fields, which has the technical scheme points of comprising a device fingerprint generation module, a parent video encryption module, a slice encryption module, a behavior recognition and defense module and an encryption logic updating and regulation module. The system generates a unique device fingerprint hash value through multi-modal feature, generates a dynamic confusion field in combination with a chaotic system and a quantum random number, realizes differential encryption of parent videos and slices, and embeds zero-knowledge blood relationship proof to support traceability verification. The behavior recognition and defense module monitors user behavior in real time, dynamically adjusts a confusion strategy or triggers an active defense mechanism, and the encryption logic updating and regulation module optimizes and updates a frequency according to playing data, so that batch crawling risks are reduced. The application can effectively improve the security and attack resistance of short video content.

Description

Technical Field

[0001] This invention relates to the field of short video encryption technology, and more specifically to a short video proactive defense encryption system based on device fingerprints and dynamic obfuscation fields. Background Technology

[0002] Device fingerprints refer to unique device characteristics or identifiers that can be used to uniquely identify a device. A device fingerprinting system is responsible for collecting user device fingerprints. Device fingerprints include information from multiple dimensions. Taking Android as an example, a device fingerprint can include the user's phone brand, Android IMEI, operating system version, Wi-Fi network identifier, memory capacity, CPU name, and so on. Collecting massive amounts of device-related information as device fingerprints serves as the raw material for risk control strategy decisions. It can be used to determine whether a user is an anomalous user, thus preparing for subsequent risk control measures.

[0003] Existing technologies for short video anti-piracy primarily rely on static obfuscation encryption, digital watermarking, and basic behavior monitoring. Static obfuscation encryption inserts fixed strings or redundant data into the video stream, but it is easily extracted and removed by reverse engineering tools. Digital watermarking embeds hidden identifiers in video footage or audio, but watermarks are easily damaged and difficult to prevent content dissemination. Basic behavior monitoring restricts access based on IP address or request frequency, but lacks effective identification capabilities for legitimate requests forged by automated tools. Furthermore, existing technologies have several limitations: obfuscation field generation rules are fixed, allowing attackers to repeatedly intercept and analyze them to crack the logic; encryption schemes are not bound to terminal device characteristics, allowing pirated content to spread across devices; relying solely on encryption and post-event traceability makes it difficult to intercept automated copyright infringement in real time; and they lack dynamic adaptability, failing to adjust protection strategies based on attack behavior, resulting in relatively simplistic defense methods. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide a short video proactive defense encryption system based on device fingerprints and dynamic obfuscation fields, thereby overcoming the aforementioned deficiencies in existing technologies.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A short video proactive defense encryption system based on device fingerprinting and dynamic obfuscation fields is characterized by comprising a device fingerprint generation module, a master video encryption module, a video segment encryption module, a behavior recognition and defense module, and an encryption logic update and control module. The device fingerprint generation module collects multimodal feature information from terminal devices and generates a unique and irreversible device fingerprint hash value using a deep neural network model. The master video encryption module uses this device fingerprint hash value and the entropy source output by a quantum random number generator to generate a master key seed through a chaotic system and constructs a dynamic obfuscation field to embed into the original video stream; the embedding position is determined by a pseudo-random function driven by the device fingerprint hash value. The video segment encryption module decomposes the encrypted master video into multiple segments, generates an independent lightweight obfuscation field for each segment, and injects a zero-knowledge lineage proof structure to support traceability verification. The behavior recognition and defense module monitors user playback behavior characteristics in real time and dynamically adjusts the obfuscation field injection strategy or triggers a proactive defense mechanism based on the risk level. The encryption logic update and control module dynamically adjusts the update frequency of the encryption logic based on the playback data of each segment, reducing the risk of being crawled in bulk.

[0007] The multimodal feature information collected by the device fingerprint generation module includes software features, hardware features, and network features. Among them, software features include operating system version number, kernel API response latency, browser rendering parameters, and trusted execution environment support flags; hardware features include CPU cache hit rate curve, GPU floating-point operation error distribution, battery charge and discharge characteristic curve, and frequency response differences of acoustic sensors; network features include DNS resolution latency distribution, Wi-Fi channel interference noise spectrum, TCP handshake time sequence, and NAT translation mode features.

[0008] The device fingerprint generation module is configured with a device fingerprint generation strategy, which includes...

[0009] Obtain the multimodal feature information.

[0010] Each of the multimodal feature information is evaluated for discriminability to obtain a corresponding feature discriminability score, which is used to characterize the degree of difference between the corresponding multimodal feature information and different terminal devices;

[0011] The preset adjustment parameters are normalized to obtain the weight parameters of each multimodal feature information. The adjustment parameters represent the degree of influence of each feature discrimination score on the weight allocation result. Among them, the multimodal feature information with a higher feature discrimination score has a lower weight parameter.

[0012] The device fingerprint hash value is then integrated and output.

[0013] The aforementioned feature information is dynamically weighted and aggregated using a federated learning framework to generate an irreversible and unique device fingerprint hash value. This module employs an adaptive weight allocation formula:

[0014]

[0015] Where D i Let represent the discriminative score of the i-th feature, α be an adjustment parameter used to control the sensitivity of weight allocation, and n be the total number of features. The final generated device fingerprint hash value has high robustness and anti-forgery capabilities.

[0016] The master video encryption module (2) is configured with a dynamic obfuscation field generation strategy, including

[0017] Obtain the device fingerprint hash value.

[0018] The device fingerprint hash value is combined with a preset chaotic model to generate a basic sequence with dynamic changing characteristics. The chaotic model is used to ensure the irregularity of the generated sequence.

[0019] Feature data fragments are extracted from the base sequence and used as the dynamic obfuscation field.

[0020] The dynamic obfuscation field generated by the master video encryption module originates from an improved Logistic chaotic system, whose core iterative formula is:

[0021] x n+1 =r·x n ·(1-x n )+β·H, where x n Let $\frac{\r}{\r}$ be the current chaotic state value, $\r$ be the chaotic system parameter, $\beta$ be the adjustment factor, and $H$ be the numerical mapping of the device fingerprint hash value. This formula combines the device fingerprint hash value with the entropy source output by the quantum random number generator to generate a variable-length binary string obfuscation field, with the field unit length ranging from 50ms to 200ms. The injection position of the obfuscation field is determined by a pseudo-random function $f(H,t)$, where $t$ is a timestamp variable, ensuring that the encrypted content generated by the same video on different devices is completely different.

[0022] The segment encryption module generates a lightweight obfuscation field for each video segment, with a length not exceeding 5% of the original segment data. It also embeds a zero-knowledge lineage proof structure generated based on the zk-SNARK protocol. This structure includes the encryption key derivation chain, device fingerprint hash value, segment index, and timestamp information, supporting verification of segment origin relationships without revealing the original key. The segment encryption process employs a layered encryption algorithm.

[0023]

[0024] Where C is the encrypted slice data, E(K,P) represents the encryption of the original slice data P using the master key seed K, and F(H,I) is the output of a pseudo-random function based on the device fingerprint hash value H and the slice index I, ensuring that the encryption logic of each slice is independent and unpredictable.

[0025] The behavior recognition and defense module includes a behavior recognition submodule, a risk assessment submodule, and a dynamic defense submodule. The behavior recognition submodule collects data such as mouse movements, click frequency, window switching behavior, and CPU usage changes during user playback, and assesses the risk level of user behavior using a support vector machine classification model. The risk assessment submodule matches preset security policies based on behavioral characteristics, while the dynamic defense submodule adjusts the obfuscation field injection strategy according to the risk level: for low-risk behaviors, the obfuscation field length is shortened to 40% of the default value; for medium-risk behaviors, fake keyframes are inserted as decoy content; and for high-risk behaviors, logic bomb code is activated, triggering abnormal termination of the player.

[0026] The encryption logic update control module dynamically adjusts the update frequency of the encryption logic based on the playback volume of each segment, specifically using an exponentially weighted moving average algorithm to calculate the update priority score.

[0027] S t =λ·V t +(1-λ)·S t-1 S t-1 Rate the current update priority as V t λ represents the current playback count of the segment, and λ is the smoothing coefficient, ranging from 0 to 1. When the rating is higher than the preset threshold, the delayed encryption logic update cycle is no less than 24 hours; when the rating is lower than the threshold, the update operation is triggered first to reduce the risk of being crawled in bulk.

[0028] The system also includes a playback behavior acquisition submodule, used to record user playback behavior data, including playback start and end times, cumulative views, concurrent playback volume, and average playback duration. This data is used to generate an encryption update priority score through a multi-dimensional clustering algorithm, driving the scheduling and update strategy of the slice encryption logic. Dynamic position function T n (f0) is used to determine the injection location of the obfuscated field, and its iterative formula is:

[0029] T n (f0)=f0+γ·H·I n

[0030] Among them, T n (f0) represents the position value after the nth iteration, f0 is the initial position, γ is the adjustment factor, and I is the slice index, ensuring that each segment of encrypted content has a differentiated position structure in different devices or playback rounds.

[0031] The zero-knowledge lineage proof structure extension binding mechanism, in addition to the encryption key derivation chain during the generation process, also binds the device fingerprint hash value, slice index, timestamp, and video unique identifier, supporting judicial-grade traceability verification. The loading path verification mechanism verifies whether the video player's slice loading order, caching strategy, and playback control behavior conform to preset policies. When abnormal behavior is detected, it triggers defensive measures such as slice freezing or encryption policy reset. The attack behavior feedback optimization module performs federated aggregation of abnormal behavior records uploaded by the user end, and reversely updates the weight parameters and obfuscation strategy strength in the federated learning model, achieving the system's self-learning defense capability against new attack paths. Attached Figure Description

[0032] Figure 1 : Schematic diagram of the device fingerprint generation module.

[0033] Figure 2 Workflow diagram of the master video encryption module.

[0034] Figure 3 : Schematic diagram of the layered encryption algorithm of the slice encryption module.

[0035] Figure 4 Functional block diagram of the behavior recognition and defense module. Detailed Implementation

[0036] This invention provides a short video proactive defense encryption system based on device fingerprinting and dynamic obfuscation fields, the specific implementation of which is described in conjunction with the accompanying drawings. Figures 1 to 4 This system comprises a device fingerprint generation module, a master video encryption module, a video segment encryption module, a behavior recognition and defense module, and an encryption logic update and control module. These modules work collaboratively to achieve efficient encryption and proactive defense of short video content. The device fingerprint generation module collects multimodal feature information from terminal devices and generates a unique and irreversible device fingerprint hash value using a deep neural network model. The master video encryption module uses this device fingerprint hash value and the entropy source output by a quantum random number generator to generate a master key seed through a chaotic system and constructs a dynamic obfuscation field to embed into the original video stream. The embedding position is determined by a pseudo-random function driven by the device fingerprint hash value. The video segment encryption module decomposes the encrypted master video into multiple segments, generates an independent lightweight obfuscation field for each segment, and injects a zero-knowledge lineage proof structure to support traceability verification. The behavior recognition and defense module monitors user playback behavior characteristics in real time and dynamically adjusts the obfuscation field injection strategy or triggers proactive defense mechanisms based on the risk level. The encryption logic update and control module dynamically adjusts the update frequency of the encryption logic based on the playback data of each segment, reducing the risk of being crawled in bulk.

[0037] First, the device fingerprint generation module collects multimodal feature information of the terminal device, including software features, hardware features, and network features. Software features include the operating system version number, kernel API response latency, browser rendering parameters, and trusted execution environment support flags; hardware features include CPU cache hit rate curves, GPU floating-point operation error distribution, battery charge / discharge characteristic curves, and frequency response differences of acoustic sensors; network features include DNS resolution latency distribution, Wi-Fi channel interference noise spectrum, TCP handshake time sequence, and NAT translation mode characteristics. For example... Figure 1 As shown, the collected multimodal feature information undergoes dynamic weight aggregation processing using a federated learning framework, employing an adaptive weight allocation formula.

[0038] Where D i Let represent the discriminative score of the i-th feature, α be an adjustment parameter used to control the sensitivity of weight allocation, and n be the total number of features. This formula generates an irreversible and unique device fingerprint hash value, ensuring that each terminal device has a unique identifier and high robustness and anti-counterfeiting capabilities.

[0039] In the master video encryption module, the master key seed, MasterSeed, is generated using the device fingerprint hash value and the entropy source output by the quantum random number generator (QRNG) through an improved Logistic chaotic system. A dynamically obfuscated field is then constructed and embedded into the original video stream. The core iterative formula of the improved Logistic chaotic system is x... n+1 =r·x n ·(1-x n )+β·H, where x n Let r be the current chaotic state value, β be the chaotic system parameter, β be the adjustment factor, and H be the numerical mapping of the device fingerprint hash value. This formula combines the device fingerprint hash value with the entropy source output by the quantum random number generator to generate a variable-length binary string confusion field, with the field unit length ranging from 50ms to 200ms. Figure 2 As shown, the injection location of the dynamically obfuscated field is determined by a pseudo-random function F(H,I), where t is a timestamp variable, ensuring that the encrypted content generated from the same video on different devices is completely differentiated. This process not only enhances the diversity of encrypted content but also significantly increases the difficulty of cracking it.

[0040] The segment encryption module decomposes the encrypted master video into multiple segments, generating an independent lightweight obfuscation field for each segment and injecting a zero-knowledge lineage proof structure to support source tracing verification. The lightweight obfuscation field is no more than 5% the length of the original segment data and embeds a zero-knowledge lineage proof structure generated based on the zk-SNARK protocol. This structure includes the encryption key derivation chain, device fingerprint hash value, segment index, and timestamp information, supporting verification of the segment's origin relationship without revealing the original key. Figure 3 As shown, the slice encryption process employs a layered encryption algorithm.

[0041] Where C represents the encrypted slice data, E(K,P) indicates that the original slice data P is encrypted using the master key seed K, and F(H,I) is the output of a pseudo-random function based on the device fingerprint hash value H and the slice index I, ensuring that the encryption logic of each slice is independent and unpredictable. Furthermore, a dynamic position function...

[0042] T n (f0)=f0+γ·H·I n ,

[0043] Among them, T n (f0) represents the position value after the nth iteration, f0 is the initial position, γ is the adjustment factor, and I is the slice index, ensuring that each segment of encrypted content has a differentiated position structure in different devices or playback rounds.

[0044] The behavior recognition and defense module includes a behavior recognition submodule, a risk assessment submodule, and a dynamic defense submodule. It monitors user playback behavior characteristics in real time and dynamically adjusts obfuscation field injection strategies or triggers proactive defense mechanisms based on risk levels. The behavior recognition submodule collects data such as mouse trajectories, click frequency, window switching behavior, and CPU usage changes during user playback, and assesses the risk level of user behavior using a support vector machine (SVM) classification model. Figure 4 As shown, the risk assessment submodule matches preset security policies based on behavioral characteristics, while the dynamic defense submodule adjusts the obfuscation field injection strategy according to the risk level: for low-risk behaviors, the obfuscation field length is shortened to 40% of the default value; for medium-risk behaviors, forged keyframes are inserted as decoy content; for high-risk behaviors, logic bomb code is activated, triggering abnormal termination of the player. The loading path verification mechanism verifies whether the video player's loading order of segments, caching strategy, and playback control behavior conform to preset policies. When abnormal behavior is detected, defensive measures such as segment freezing or encryption policy reset are triggered.

[0045] The encryption logic update control module dynamically adjusts the update frequency of the encryption logic based on the playback volume of each segment, specifically using the Exponentially Weighted Moving Average (EWMA) algorithm to calculate the update priority score. The update priority score formula is S.t =λ\V t +(1-λ)\S t-1 S t Rate the current update priority as V t The current segment's playback count is represented by λ, a smoothing coefficient ranging from 0 to 1. When the score exceeds a preset threshold, the encryption logic update cycle is delayed by at least 24 hours; when the score falls below the threshold, an update operation is triggered first to reduce the risk of being crawled in bulk. The playback behavior collection submodule records user playback behavior data, including playback start and end times, cumulative views, concurrent playback counts, and average playback duration. This data is used to generate an encryption update priority score through a multi-dimensional clustering algorithm, driving the scheduling and update strategy of the segment encryption logic.

[0046] The attack behavior feedback optimization module performs federated aggregation of abnormal behavior records uploaded by users, and reversely updates the weight parameters and obfuscation strategy strength in the federated learning model, enabling the system to achieve self-learning defense against new attack paths. The zero-knowledge lineage proof structure extension binding mechanism, in addition to the encryption key derivation chain during generation, also binds the device fingerprint hash value, slice index, timestamp, and video unique identifier, supporting judicial-level source tracing verification. This design not only effectively prevents the spread of unauthorized content but also provides a complete chain of evidence to support legal accountability when necessary. The loading path verification mechanism verifies whether the video player's slice loading order, caching strategy, and playback control behavior conform to preset strategies. When abnormal behavior is detected, it triggers defensive measures such as slice freezing or encryption strategy reset. The attack behavior feedback optimization module performs federated aggregation of abnormal behavior records uploaded by users, and reversely updates the weight parameters and obfuscation strategy strength in the federated learning model, enabling the system to achieve self-learning defense against new attack paths.

[0047] In summary, this invention achieves full lifecycle protection for short video content through the coordinated operation of a device fingerprint generation module, a master video encryption module, a slice encryption module, a behavior recognition and defense module, and an encryption logic update and control module. From device fingerprint generation to the construction of dynamic obfuscation fields, and then to slice encryption and behavior monitoring, the entire system demonstrates excellent security and flexibility in both technical aspects and practical applications. As can be seen from the above specific implementation methods, this system not only effectively counters traditional attack methods but also possesses strong adaptability and scalability, making it widely applicable to various short video platforms and content distribution scenarios, providing robust protection for users and content creators.

[0048] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A short video proactive defense encryption system based on device fingerprinting and dynamic obfuscation fields, characterized in that: include The device fingerprint generation module collects multimodal feature information of the terminal device and generates a device fingerprint hash value. The master video encryption module uses the device fingerprint hash value and a chaotic model to generate a dynamic obfuscation field, which is then embedded into the original video stream. The embedding position is determined by a dynamic position function driven by the device fingerprint hash value, thus forming an encrypted master video. The segment encryption module decomposes the encrypted master video into multiple segments, generates an independent segment obfuscation field for each segment, and embeds it into the segment video stream to form encrypted segment videos. Simultaneously, it embeds an association proof, which indicates the association relationship between the encrypted segment videos and the encrypted master video. The behavior recognition and defense module monitors user playback behavior characteristics in real time and dynamically adjusts the obfuscation field injection strategy or triggers an active defense mechanism based on the risk level. The encryption logic update control module dynamically adjusts the update frequency of the segment obfuscation field based on the playback data of each segment.

2. The short video proactive defense encryption system based on device fingerprint and dynamic obfuscation field according to claim 1, characterized in that: The device fingerprint generation module collects multimodal feature information including software feature information, hardware feature information, and network feature information. Based on a federated learning model, the multimodal information is aggregated according to weights to obtain the device fingerprint hash value. The software feature information includes operating system version number, kernel API response latency, browser rendering parameters, and trusted execution environment support flags. The hardware feature information includes CPU cache hit rate curve, GPU floating-point operation error distribution, battery charge and discharge characteristic curve, and acoustic sensor frequency response differences. The network feature information includes DNS resolution latency distribution, Wi-Fi channel interference noise spectrum, TCP handshake time sequence, and NAT translation mode characteristics.

3. The short video proactive defense encryption system based on device fingerprint and dynamic obfuscation field according to claim 2, characterized in that: The device fingerprint generation module is configured with a device fingerprint generation strategy, which includes... Obtain the multimodal feature information. Each of the multimodal feature information is evaluated for discriminability to obtain a corresponding feature discriminability score, which is used to characterize the degree of difference between the corresponding multimodal feature information and different terminal devices; The preset adjustment parameters are normalized to obtain the weight parameters of each multimodal feature information. The adjustment parameters represent the degree of influence of each feature discrimination score on the weight allocation result. Among them, the multimodal feature information with a higher feature discrimination score has a lower weight parameter. The device fingerprint hash value is then integrated and output.

4. The short video proactive defense encryption system based on device fingerprint and dynamic obfuscation field according to claim 1, characterized in that: The master video encryption module is configured with a dynamic obfuscation field generation strategy, including... Obtain the device fingerprint hash value. The device fingerprint hash value is combined with a preset chaotic model to generate a basic sequence with dynamic changing characteristics. The chaotic model is used to ensure the irregularity of the generated sequence. Feature data fragments are extracted from the base sequence and used as the dynamic obfuscation field.

5. The short video proactive defense encryption system based on device fingerprint and dynamic obfuscation field according to claim 1, characterized in that: The slice encryption module is configured with an associated proof embedding strategy, including... Obtain the encryption key derivation link corresponding to the encrypted video slice, where the encryption key derivation link represents the derivation relationship between the slice encryption key and the parent video encryption key. The device fingerprint hash value, the slice index corresponding to the current segment, and the timestamp when the slice was generated are obtained, and a lineage proof structure is generated. The lineage proof structure is used to verify the association between the encrypted slice video and the encrypted parent video. Embed the bloodline proof structure into the segmented video stream.

6. The short video proactive defense encryption system based on device fingerprint and dynamic obfuscation field according to claim 5, characterized in that: When the slice encryption module encrypts each slice using a layered encryption algorithm, it first obtains the original slice data; then, it encrypts the original slice data using core encryption information to obtain the first layer of encrypted data; simultaneously, based on the device fingerprint hash value generated by the device fingerprint generation module and the slice index corresponding to the current slice, it generates exclusive random data corresponding to the slice through a pseudo-random function; the first layer of encrypted data and the exclusive random data are then fused to obtain the encrypted slice data, so that the encrypted slice data is associated with the core encryption information, the device fingerprint hash value, and the slice index, thereby achieving layered encryption protection for the slice.

7. The short video proactive defense encryption system based on device fingerprint and dynamic obfuscation field according to claim 1, characterized in that: The encryption logic update control module is used to dynamically adjust the update frequency of the slice obfuscation field corresponding to each segment. Its control logic is as follows: the update frequency is determined based on the number of views of each segment. The higher the number of views of a segment, the lower the update frequency is set to reduce the risk of the update logic being illegally crawled due to frequent updates. The update frequency of segments with lower views can be relatively increased. At the same time, the encryption logic update control module is associated with the dynamic obfuscation field adjustment mechanism of the parent video encryption module. When the parent video encryption module makes a fixed adjustment to the dynamic obfuscation field, the slice obfuscation field of each segment is updated synchronously with the adjustment to maintain the dynamism and relevance of the encryption logic.

8. The short video proactive defense encryption system based on device fingerprint and dynamic obfuscation field according to claim 1, characterized in that: The behavior recognition and defense module includes a behavior recognition submodule, a risk assessment submodule, and a dynamic defense submodule. The behavior recognition submodule collects the user's mouse trajectory, click frequency, window switching behavior, and CPU usage changes during playback and assesses the user's behavior risk level using a support vector machine classification model.

9. The short video proactive defense encryption system based on device fingerprint and dynamic obfuscation field according to claim 8, characterized in that: The dynamic defense submodule adjusts the obfuscation field injection strategy according to the risk level. Under low-risk behavior, it shortens the obfuscation field length to 40% of the default value. Under medium-risk behavior, it inserts fake keyframes as decoy content. Under high-risk behavior, it activates logic bomb code to trigger abnormal termination of the player.

10. The short video proactive defense encryption system based on device fingerprint and dynamic obfuscation field according to claim 5, characterized in that: The lineage proof structure is a zero-knowledge lineage proof structure, and in the generation process, in addition to the encryption key derivation link, it is also bound to the device fingerprint hash value, slice index, timestamp and video unique identifier.

Images