Mechanisms are provided for providing a hardened neural network. The mechanisms configure the hardened neural network executing in the data processing system to introduce noise in internal feature representations of the hardened neural network. The noise introduced in the internal feature representations diverts gradient computations associated with a loss surface of the hardened neural network. The mechanisms configure the hardened neural network executing in the data processing system to implement a merge layer of nodes that combine outputs of adversarially trained output nodes of the hardened neural network with output nodes of the hardened neural network trained based on the introduced noise. The mechanisms process, by the hardened neural network, input data to generate classification labels for the input data and thereby generate augmented input data which is output to a computing system for processing to perform a computing operation.

Protecting cognitive systems from gradient based attacks through the use of deceiving gradients

The invention discloses a text sentiment classification method based on a deep learning combined model. The method comprises the steps of (1) training a word vector and a character vector; (2) performing word segmentation for each sentence of an annotated text, filling to a fixed length to obtain a training dataset I, performing character segmentation on each sentence of the annotated text, filling to the fixed length to obtain a training dataset II; (3) endowing words and characters of the two training datasets with corresponding word vectors and character vectors; (4) training the two training datasets by use of a text CNN (Convolutional Neural Network) and an improved Dynamic CNN model to obtain four base classifiers, wherein a progressive learning method is adopted as a training methodand a focal loss function is adopted as a loss function and (5) combining the four base classifiers to obtain the text sentiment classification method of a combined model. The text sentiment classification method implemented by the invention does not depend on a specific sentiment dictionary, is not limited to a specific field and has high universality and expandability.

Text sentiment classification method based on deep learning combined model

The invention discloses a confrontation sample generation method oriented to Chinese text emotion tendency detection. Firstly, the input sample data is preprocessed. Then the Chinese word dictionary is constructed, and the substitute model of depth learning is constructed, and the training and parameter adjustment of the model are carried out. Finally, input the normal sample into the depth learning substitution model, calculate the importance of each word in the input, and modify it to generate antagonistic sample. The semantic change of the sample data to the original data is slight, which does not affect the understanding of the meaning of the expression. The invention proves the existence of antagonistic attack based on Chinese information, verifies that the Chinese antagonistic samplecan induce the deep neural network model to generate wrong classification, and further proves the mobility of the antagonistic sample. This paper reveals the mechanism of text types against sample attacks, and provides examples for the security verification and testing of deep neural network model against sample attacks, and provides data sources for defensive countermeasure training methods.

Method of antagonistic sample generation for Chinese text affective orientation detection

Mechanisms are provided for obfuscating training of trained cognitive model logic The mechanisms receive input data for classification into one or more classes in a plurality of predefined classes as part of a cognitive operation of the cognitive system The input data is processed by applying a trained cognitive model to the input data to generate an output vector having values for each of the plurality of predefined classes A perturbation insertion engine modifies the output vector by inserting a perturbation in a function associated with generating the output vector, to thereby generate a modified output vector The modified output vector is then output The perturbation modifies the one or more values to obfuscate the trained configuration of the trained cognitive model logic while maintaining accuracy of classification of the input data

Protecting cognitive systems from model stealing attacks

The invention provides a three-class emotion recognition model method based on the convolutional neural network, comprising: acquiring an electroencephalogram signal generated by a participant in emotional changing, and constructing a three-class emotion recognition model based on the convolutional neural network; using the acquired electroencephalogram signal to train the constructed three-classemotion recognition model based on the convolutional neural network; using the trained three-class emotion recognition model based on the convolutional neural network to recognize a new electroencephalogram signal generated by the participant. An acquired signal is down-sampled based on electroencephalogram to obtain a discrete electroencephalogram signal; a three-class emotion recognition model based on the convolutional neural network is constructed; finally the three-class emotion recognition model based on the convolutional neural network is trained to obtain the three-class emotion recognition model based on the convolutional neural network which is suitable for emotion recognition. The three-class emotion recognition model based on the convolutional neural network can provide improved robustness and accuracy for emotion recognition results.

Three-class emotion recognition model method based on convolutional neural network

The invention discloses an identification method of emotional tendency of network comment texts and a convolutional neutral network model. The method comprises the steps as follows: grabbed network comment texts constitute a data set; word segmentation and text preprocessing are performed; all words subjected to text preprocessing are trained, and word vector representation of all words is obtained; the convolutional neutral network model is constructed and trained on a training set selected from the data set, and network parameters are updated with a back-propagating algorithm; in each step of training, noise is added to word vectors of an input layer for construction of adversarial samples, adversarial training is performed, and network parameters are updated with a random gradient descent algorithm; a classification model is obtained through repeated iteration to identify the emotional tendency of the network review texts. The convolutional neutral network model is used in the method and comprises the input layer, a convolution layer, a pooling layer and a classification layer. The adversarial samples can be classified correctly and the identification accuracy is improved.

Identification method of emotional tendency of network comment texts and convolutional neutral network model

The invention discloses a user identity authentication method suitable for a transparent computing system and a system configuration method. The user identity authentication method comprises: when a user boots up and logs in, a client transmits a login requesting information to an authentication server, and the authentication server queries a background database, acquires an operating system authority list bootable by the user and transmits the list to the client; the user selects the operating system and invokes a remote loading request to a storage server; and the storage server authenticates a user name and passwords of the client, checks a mirror image configuration file of the operating system of the user after the authentication passes, and judges whether the authentication succeeds according to the IP of the client and mirror image IP settings of an operating system to be booted up. The system configuration method is used for configuring the transparent computing system using the user identity authentication method. According to the user identity authentication method suitable for the transparent computing system and the system configuration method, which are provided by the invention, the authority management of the user is achieved, a plurality of individual services can be provided, and the user experience is perfected.

User identity authentication method suitable for transparent computing system and system configuration method

The invention discloses a making method of an operating system mirror image applicable to iSCSI protocol remote wireless loading. The method includes: configuring iSCSI service at a server end, installing an operating system for remote loading, modifying the RAMDISK virtual file system of an instance operating system of a terminal with the loaded instance operating system, and driving a wireless network card in the RAMDISK virtual file system to connect to an appointed wireless network so as to complete the wireless loading process of the operating system mirror image through an iSCSI protocol. By the method, the difficulty that the operating system mirror image cannot be remotely wirelessly loaded through the iSCSI protocol is solved, and the application range of the operating system remote loading technology is expanded.

Zheng Jin

Papers

Identification method of emotional tendency of network comment texts and convolutional neutral network model

User identity authentication method suitable for transparent computing system and system configuration method

Making method of operating system mirror image applicable to iSCSI protocol remote wireless loading