7 results about "Obfuscation" patented technology

The present disclosure relates to systems, non-transitory computer-readable media, and methods for utilizing machine learning to generate an obfuscated summary for employee feedback data and generating a modification suggestion based on the employee feedback data. In particular, in one or more embodiments, the disclosed systems generate a prompt for an obfuscation and summary generation large language model to generate an obfuscated summary of employee feedback data and provide the obfuscated summary within a manager feedback interface on a manager client device. Moreover, in one or more embodiments, the disclosed systems receive a request to generate a modification suggestion from the manager feedback interface, then utilize a recommendation large language model to generate a modification suggestion based on the employee feedback data. Further, the disclosed systems provide the modification within the manager feedback interface on the manager client device.

The invention discloses a code obfuscation method, a business execution method, a medium, equipment and a product. According to the scheme, the to-be-mixed first language code and the to-be-mixed second language code are firstly obtained, then the second language code serves as the character string constant in the first language code and is written into the first language code, the mixed code is obtained, finally, the mixed code is subjected to obfuscation processing through the preset security compiler, and the safety of the mixed code is improved. And obtaining an executable file corresponding to the mixed code.

The application discloses a method for de-obfuscation and feature extraction of phishing documents and application thereof in attack detection, wherein the de-obfuscation comprises obtaining obfuscated macro code of the phishing document, constructing a hint engineering template by using a pre-trained language model, analyzing an obfuscated logic structure and generating de-obfuscation rules and restoration strategies; the obfuscated macro code is structured into an abstract syntax tree by using an analysis tool, typical obfuscated patterns are matched based on a regular expression, function execution simulation and cell reference analysis are combined, preliminary restoration of the structure, restoration of control flow semantics and operation path construction are realized; according to the de-obfuscation rules and the abstract syntax tree, the obfuscated structure is converted into readable macro statements, a macro code instruction sequence without obfuscated semantics is generated, and a semantic sequence after de-obfuscation is output. The feature extraction comprises extraction of word features, Token features, abstract syntax tree features and relationship features. The application breaks through the bottleneck of traditional phishing attack detection and difficulty in identification of obfuscated documents, and improves the detection accuracy of phishing document attacks.

This disclosure provides a code obfuscation method applicable to the field of information security technology. The method includes: obtaining a first function to be obfuscated and a second function to be obfuscated; performing a first obfuscation process on the first function to be obfuscated to obtain a first code fragment, wherein the first obfuscation process includes rewriting the function content and call relationships of the first function to be obfuscated; performing a second obfuscation process on the second function to be obfuscated to obtain a second code fragment, wherein the second obfuscation process includes rewriting the call relationships of the second function to be obfuscated; and obtaining obfuscated code after the first and second functions to be obfuscated have completed their respective obfuscation processes. This disclosure also provides a code obfuscation apparatus, device, storage medium, and program product.

A method and apparatus for multi-party joint fine-tuning of a language model based on data protection are disclosed. The multi-party includes a first party holding a target language model and a second party providing computing power. The target language model includes an embedding table and serially configured multi-layer network layers. The first party determines multi-layer obfuscation layers and an obfuscation embedding table; at least the multi-layer obfuscation layers are sent to the second party. The obfuscation embedding table is obtained by performing a second obfuscation on the embedding table, and the obfuscation layers are obtained by performing a first obfuscation on the parameter matrices in the corresponding network layers. The first obfuscation method corresponds to the second obfuscation method. The second party obtains training samples including the target obfuscated word embedding sequence and its labels, which are determined based on the text and the obfuscation embedding table. Based on the target obfuscated word embedding sequence, a first output result is obtained through the multi-layer obfuscation layers. Based on the difference between the first output result and the labels, the multi-layer obfuscation layers are adjusted to achieve multi-party joint fine-tuning of the model while protecting model parameters and data privacy.

The invention relates to the technical field of speech translation, in particular to a cross-type combined transport confusion method and device based on adaptive optimal transmission. The method comprises the following steps: constructing an optimal transmission model of a multi-task universal framework; optimal transmission alignment based on attention enhancement is carried out on the voice and text sequences; on the basis of a dynamic window strategy, optimal transmission alignment based on attention enhancement is optimized, and the alignment relation between the voice and the text sequence is obtained; fusing the voice and text features of the voice and text sequences through a similarity-based adaptive mixing strategy; and combining a contrast learning loss function and a multi-task learning framework, and combining a multi-task loss function by constructing positive and negative sample pairs to obtain a unified optimization target. Through the dynamic window strategy, the optimal transmission, the comparative learning and other technologies, the representation difference between the voice and the text is effectively reduced, and the translation task of the low-resource language is remarkably improved.

This application relates to the field of network security technology and discloses a method, apparatus, and device for deobfuscating JavaScript obfuscated code. The method includes: identifying and extracting obfuscation feature data and core code from the obfuscated JavaScript code to be processed; segmenting the core code based on a first threshold to obtain multiple core code fragments with a length less than or equal to the first threshold; deobfuscating each core code fragment based on the obfuscation feature data using a first large language model to obtain a corresponding first code fragment; extracting a backbone code fragment from each first code fragment using a second large language model and fusing all backbone code fragments to obtain a target backbone code; and filling the function body code corresponding to each backbone code fragment into the target backbone code to generate the target JavaScript code. This method can achieve automated deobfuscation of long JavaScript code, improving the success rate and correctness of deobfuscating long text code.