Search

CN-122024746-A - Hiding method for browser audio fingerprints

CN122024746ACN 122024746 ACN122024746 ACN 122024746ACN-122024746-A

Abstract

The invention discloses a hiding method of browser audio fingerprints, which relates to the technical field of network security, and comprises the steps of capturing browser audio data in real time, extracting key parameters such as length, frequency, amplitude and the like and spectrum distribution characteristics, constructing a standardized audio characteristic data set, screening audio non-key fragments, generating self-adaptive noise matched with spectrum characteristics and dynamically injecting the self-adaptive noise, realizing characteristic reconstruction through phase disturbance and timestamp migration, encrypting virtual audio characteristics by combining a hash salifying algorithm of dynamic salt values, verifying whether the audio distortion degree meets a threshold value and the avoidance recognition effect of the virtual characteristics through simulating a fingerprint acquisition process, optimizing related parameters by adopting a gradient descent method based on a verification result, storing optimal combinations, self-adaptively calling optimal parameters according to webpage types and historical data, and finally clearing session parameters and processing a buffer area after audio playing is completed, thereby effectively guaranteeing user privacy.

Inventors

  • SHANG SHUANGSHUANG

Assignees

  • 江苏灵匠信息科技有限公司

Dates

Publication Date
20260512
Application Date
20260324

Claims (10)

  1. 1. A method for hiding an audio fingerprint of a browser, comprising the steps of: S1, capturing audio data output by a browser in real time, extracting length, frequency and amplitude parameters of the audio data as key parameters, simultaneously acquiring spectrum distribution characteristics of an audio signal through a spectrum analysis tool, and generating an audio characteristic data set based on the key parameters and the spectrum distribution characteristics; S2, dividing the audio data into a plurality of fragments according to a preset rule based on the audio type and the time interval, screening out non-key fragments as interference objects through an audio content identification algorithm, calling a noise intensity adjustment algorithm to generate self-adaptive noise matched with the spectrum distribution characteristics, and injecting the self-adaptive noise into the selected fragments according to a dynamic proportion; S3, carrying out feature reconstruction on the interfered audio data based on a preset fingerprint variation rule, generating virtual audio features which are not related to the original audio fingerprint by modifying phase parameters and time stamp information of the audio signals, and simultaneously carrying out encryption processing on the virtual audio features by utilizing a hash salting algorithm; S4, simulating a website audio fingerprint acquisition process, respectively acquiring and comparing the audio fingerprints before and after interference and confusion, detecting whether the audio playing effect meets the distortion threshold requirement, and simultaneously verifying whether the encrypted virtual audio features can avoid the identification of a fingerprint analysis tool; S5, according to the verification result, taking the avoidance success rate as a target of more than or equal to 95%, dynamically adjusting the fragment selection range in the step S2, the noise intensity of the self-adaptive noise and the characteristic reconstruction parameters and encryption parameters in the step S3 by adopting a gradient descent method, generating an optimal interference and confusion parameter combination, and storing the optimal interference and confusion parameter combination in a parameter library; S6, based on the webpage types accessed by the user, the historical interference verification data and the historical parameter calling records, the corresponding optimal parameter combination in the parameter library is automatically called, and self-adaptive interference adjustment is achieved.
  2. 2. The method for hiding the browser audio fingerprint according to claim 1, wherein in the step S1, the construction of the standardized audio feature data set includes respectively carrying out normalization processing on the length, frequency and amplitude parameters by using a min-max normalization algorithm, uniformly mapping the parameter values to the [0,1] intervals, and retaining the original distribution trend of the parameters.
  3. 3. The method for hiding the audio fingerprint of the browser according to claim 1, wherein in the step S1, a frequency interval of the spectrum analysis tool is divided into 16-64 frequency subintervals by adopting an equal frequency interval strategy, wherein the audio with rich high frequency details is divided into 32-64 subintervals, and the audio with less high frequency details is divided into 16-32 subintervals.
  4. 4. The method of hiding an audio fingerprint of a browser as set forth in claim 1, wherein in step S3, according to a preset random seed, the phase parameters of the audio signal are randomly disturbed in segments according to the audio time segments, and random offsets subject to normal distribution are added to the timestamp information of the audio data to generate virtual audio features unrelated to the original audio fingerprint.
  5. 5. The method for hiding the audio fingerprint of the browser according to claim 1, wherein in step S3, the salt value used by the hash salifying algorithm is dynamically generated, the hash value is generated according to the system timestamp and the audio data fragment, and the salt value is updated every time the virtual audio feature is generated.
  6. 6. The method for hiding the audio fingerprint of the browser according to claim 1, wherein in the step S4, the distortion threshold includes a signal-to-noise ratio threshold and a spectral correlation threshold, wherein the signal-to-noise ratio threshold is not lower than 25dB, the spectral correlation threshold is not lower than 0.85, and when the audio playing effect is detected, the signal-to-noise ratio and the spectral correlation coefficient of the audio signal before and after the interference confusion are calculated respectively, and only when both indexes meet the corresponding threshold requirements, the verification is determined.
  7. 7. The method for hiding browser audio fingerprints according to claim 1, wherein in step S4, at least two different browser audio fingerprint acquisition scripts are adopted for acquisition and comparison analysis when simulating a website audio fingerprint acquisition flow.
  8. 8. The method for hiding the browser audio fingerprint according to claim 1, wherein in step S5, the process of dynamically adjusting is performed by iteratively optimizing the noise intensity, the segment selection range, the feature reconstruction parameter and the encryption parameter by using a gradient descent method with the avoidance success rate as a target according to the verification result of step S4 until the avoidance success rate target or the maximum iteration number is reached.
  9. 9. The method of hiding audio fingerprints of a browser according to claim 1, wherein in step S6, the types of web pages accessed by the user are classified according to the domain name of the web page and the calling frequency of the audio class API in the web page, the calling frequency is higher than 5 times/min for the high-frequency audio class web page, the calling frequency is not higher than 5 times/min for the low-frequency audio class web page, and the corresponding optimal parameter combination is matched and called from the parameter library according to the type label of the web page.
  10. 10. The method for hiding an audio fingerprint of a browser as recited in claim 1, further comprising step S7 of clearing all noise intensity parameters, segment selection parameters, feature reconstruction parameters and encryption parameters generated in the current session after audio playback is completed, and writing random white noise data in an audio buffer of the browser.

Description

Hiding method for browser audio fingerprints Technical Field The invention relates to the technical field of network security, in particular to a hiding method of an audio fingerprint of a browser. Background With the rapid development of internet technology, a browser is a core entrance for users to access the network world and develop various online activities, in the process, the browser is used as an interaction medium between user equipment and network services, the related characteristic information of the equipment and the users can be unconsciously exposed, audio fingerprints are one of key equipment identification technologies, and the generation principle of the audio fingerprints of the browser is to utilize inherent differences among different equipment hardware, operating system audio drivers and browser audio processing engines. The website plays a preset audio signal or induces equipment to generate audio data by calling relevant interfaces such as WebAudioAPI of a browser, and then carries out processing such as spectrum analysis, parameter extraction and the like on the returned audio signal, so that a unique equipment audio fingerprint which can be stably identified is finally formed. However, the unaware equipment tracking behavior infringes the privacy rights of users, sensitive information is illegally collected, analyzed and even traded under the condition that the users are unaware, the existing anti-audio fingerprint tracking technology is mainly focused on modes of limiting browser audio API call authority, directly shielding an audio data acquisition request or returning false audio data in a fixed format and the like, but the methods have obvious defects that limiting API call or shielding the acquisition request can cause that part of audio functions of websites cannot be normally used, normal surfing experience of the users is influenced, the mode of returning the fixed false audio data is easy to be identified and marked as 'abnormal fingerprint' by a tracking algorithm, and then the website refuses service or adopts a stricter tracking strategy, and the method cannot adapt to the audio fingerprint acquisition algorithm which is continuously iteratively updated, so that the anti-tracking stability and compatibility are poor. Disclosure of Invention The invention aims to provide a hiding method for audio fingerprints of a browser, which aims to solve the problems in the background technology. In order to achieve the above purpose, the invention provides a method for hiding an audio fingerprint of a browser, which comprises the following steps: S1, capturing audio data output by a browser in real time, extracting length, frequency and amplitude parameters of the audio data as key parameters, simultaneously acquiring spectrum distribution characteristics of an audio signal through a spectrum analysis tool, and generating an audio characteristic data set based on the key parameters and the spectrum distribution characteristics; S2, dividing the audio data into a plurality of fragments according to a preset rule based on the audio type and the time interval, screening out non-key fragments as interference objects through an audio content identification algorithm, calling a noise intensity adjustment algorithm to generate self-adaptive noise matched with the spectrum distribution characteristics, and injecting the self-adaptive noise into the selected fragments according to a dynamic proportion; S3, carrying out feature reconstruction on the interfered audio data based on a preset fingerprint variation rule, generating virtual audio features which are not related to the original audio fingerprint by modifying phase parameters and time stamp information of the audio signals, and simultaneously carrying out encryption processing on the virtual audio features by utilizing a hash salting algorithm; S4, simulating a website audio fingerprint acquisition process, respectively acquiring and comparing the audio fingerprints before and after interference and confusion, detecting whether the audio playing effect meets the distortion threshold requirement, and simultaneously verifying whether the encrypted virtual audio features can avoid the identification of a fingerprint analysis tool; S5, according to the verification result, taking the avoidance success rate as a target of more than or equal to 95%, dynamically adjusting the fragment selection range in the step S2, the noise intensity of the self-adaptive noise and the characteristic reconstruction parameters and encryption parameters in the step S3 by adopting a gradient descent method, generating an optimal interference and confusion parameter combination, and storing the optimal interference and confusion parameter combination in a parameter library; S6, based on the webpage types accessed by the user, the historical interference verification data and the historical parameter calling records, the corresponding optimal parameter combinat