30 results about "Information sensitivity" patented technology

Described is a mechanism by which application programs (or plug-ins) and an auxiliary service adjust the output and / or data available for output on an auxiliary device based on changes detected in the current context of the auxiliary display. Context sensors detect and report changes in context, whereby some action is taken with respect to what information is displayed or can be displayed on the auxiliary display device. Examples of context data that can change include the physical location of the display, detected movement or motion data, presence of a particular user or others, power state, network connectivity status, privacy status and so forth. The auxiliary device can have context as to its position relative to the main display. Information sensitivity levels may be used to specify whether an application program's data can be displayed, based on the perceived sensitivity of the information and the user's current context.

A device for analyzing reputation for objects by data mining is provided. The device includes a big data collector assembly collecting first information from a web page, a text preprocessor assembly analyzing morpheme information from an original text of the first information, an ontology dictionary extender assembly classifying the morpheme information according to each category and extracting an associated word, an object recognizer assembly measuring similarity between the associated word from the morpheme information to recognize a plurality of entity names associated with a first object, an information sensitivity classifier assembly classifying sensitivity of the first information based on a frequency of a sensitive word associated with the first object from the morpheme information, and a risk evaluation modeler assembly deriving a numeric risk value of the first information according to a numeric sensitivity value of the sensitive word and a weight value level for evaluating event importance.

The invention discloses an image significance object detection method based on multiscale low-rank decomposition and with sensitive structural information. The method comprises the following steps of: a three-dimensional volume data generation stage: performing superpixel decomposition and DT (Delaunay Triangulation) on a two-dimensional image to form three-dimensional volume data corresponding to the two-dimensional image, a Biharmonic distribution calculation stage: obtaining a Biharmonic diffusion result of each superpixel point, a subgeneration description stage: performing histogram statistics on L2 distances between sampling points on Biharmonic isolines of the superpixel points to form shape description of the isolines, and a multiscale low-rank decomposition stage: obtaining a final significance object detection result by differencing sparse matrixes obtained by the low-rank decomposition under different scales and performing residual error matrix summation based on the shape description of the Biharmonic isoline of each superpixel point. The method is based on GPU (Graphics Processing) parallel implementation, can detect one or more significance objects in the image, and has the characteristics of high detection precision of the significance object, complete detection of the significance object, good noise resistance and the like.

The invention discloses a sensitive long-term and short-term memory method based on input variation differentiation. In order to improve the traditional LSTM neural network's ability to respond to short-term information, a neural unit of the long-term and short-term memory network with increased information sensitivity is added, which can It greatly increases its ability to respond to short-term information, improves the real-time performance of its application, and enables more complete real-time analysis, further analysis of micro-actions and other contents, and improves the application value.

There is described a method for determining a level of sensitivity of information in an electronic document. The method comprises scanning a computer location to select the electronic document, such as an unstructured document in which the sensitive nature of a given portion of the contents is not trivial. In the electronic document, contents and metadata of the electronic document are scanned, and each occurrence of sensitive data is identified by classifying each portion of the contents forming the electronic document as sensitive, or not sensitive, per se. For each occurrence of the sensitive data, there are determined a type of the sensitive data and a risk score associated to the type of the sensitive data, for example from a knowledge base. Using the risk score of each occurrence of the sensitive data, one can determine an exposure risk score of the electronic document.

The embodiment of this specification provides a method for protecting the safety of a neural network model, including: obtaining a neural network model, including multiple network layers obtained by training with training data; for any first network layer among them, fixing other network layer parameters In the case of , the above-mentioned training data is used to perform the first parameter adjustment on the first network layer to obtain the first fine-tuning model; determine the first index value corresponding to the preset performance index of the first fine-tuning model, and the index of the preset performance index The value depends on the corresponding model, the relative size between the test loss on the test data and the training loss on the above training data; similarly, using the training data and test data to perform the second parameter adjustment on the first network layer, we get The second fine-tuning model, and determine the second index value; based on the relative size of the first index value and the second index value, determine the information sensitivity corresponding to the first network layer, and if it is greater than a predetermined threshold, the first network layer for security processing.

The invention discloses a sensitive long-term and short-term memory method based on state change differentiation. In order to improve the traditional LSTM neural network's ability to respond to short-term information, a neural unit of the long-term and short-term memory network with increased information sensitivity is added, which can It greatly increases its ability to respond to short-term information, improves the real-time performance of its application, and enables more complete real-time analysis, further analysis of micro-actions and other contents, and improves the application value.