The bad text containing variant words seriously harms the health of the network environment. The existing methods for the recognition of bad variant text do not take into account the importance of the phonetic and positional information for the recognition of bad variant text. In this paper, a ChineseBERT-BiGRU variant bad text recognition model is proposed. This model first learns and trains the word vector of the text by inputting the pinyin information, font information and character information of the text, and then combines the position information of the text. Then the word vector of the text is input into BiGRU to learn richer semantic information of the text. Finally, the bad text containing variant words is identified through Softmax classification. The accuracy, accuracy and F1 values of ChineseBERT-BiGRU in the data set in this paper are compared with those of other models.
With the increasing popularity of the Internet, network security issues are also increasing and there are more and more fraudulent web pages, which also brings great obstacles to the governance of network security. However, for the detection of fraudulent web pages, most of the previous detection methods are based on web page characteristics. Multiple information on web page content, such as web page text, HTML, images, and other content is also an important feature for detecting and discovering fraudulent web pages. In this paper, the text dataset of fraudulent web pages obtained by the crawler technique is used to extract the features of the text content sentences using the bidirectional LSTM model, and the feature weights of the text dataset of fraudulent web pages are enhanced by introducing an attention mechanism to obtain a feature vector representation of the text data sentences and then the classification detection is carried out. The experiments show that the accuracy and F1 values of the bidirectional LSTM model are higher than those of other models. The bidirectional LSTM model on the addition of the attention mechanism performed better, yielding an accuracy result of 90.85%, which is a 2.7% increase in accuracy and 3.09% increase in F1 value compared to the bidirectional LSTM model. Therefore, the method is effective in detecting fraudulent web pages and has some practicality in cyber security governance.
Cybersecurity event extraction aims to extract threat intelligence information from unstructured text and provide a data basis for correlation analysis of cybersecurity events. However, in the field of cybersecurity, the complexity of terminology in event data as well as the problem of polysemy and crossover between Chinese and English in Chinese texts pose great challenges to event extraction. Existing methods usually perform event extraction in a pipelined manner, ignoring the dependencies between event elements and event triggers. Therefore, a joint event extraction model is proposed in this paper. The model captures the contextual features of sentences in the encoding layer by an improved temporal convolutional network (TCN) and then enhances the dependency features between triggers and arguments using a multihead attention mechanism to achieve sentence-level joint extraction of Chinese cybersecurity events. We conducted comparison experiments on real cybersecurity news data to evaluate the event extraction performance of the model. Compared with a baseline model LSTM, the method improves the F1 values by 13.2% and 33.2% in two subtasks of trigger extraction and argument extraction, respectively.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.