The rapid proliferation of Internet of Things (IoT) devices has led to a significant increase in the number of network users, prompting advancements in security mechanisms. Consequently, traditional attacks targeting specific vulnerabilities have become less effective due to these enhanced defense systems, leading attackers to increasingly adopt phishing strategies as a primary means of bypassing security measures. Among these, phishing websites have been increasing rapidly, exploiting the carelessness of countless users. In response, numerous phishing website detection methods have been investigated, with machine learning-based approaches emerging as a leading strategy. However, these machine learning-based classification methods require substantial computational resources, posing challenges for their direct application in the already widespread IoT environment. To address these challenges, we propose an efficient phishing website detection method based on HTML tag sequences, the core structural elements of websites, by leveraging encoder models known for their effectiveness in classifying sequential data. Our approach also incorporates a customized tokenizer and dictionary specifically tailored for HTML tags. Experiments conducted on publicly available datasets demonstrate that the proposed method achieves over 95% accuracy across key performance metrics. Furthermore, comparative analyses highlight several advantages of our method, including reduced model size and faster detection times compared to existing approaches.

Many prediction models have been proposed to improve the effectiveness of web prefetching for reducing the response time perceived by users when browsing the web. Most of these models are based on structure learning and are applied at the client side. Currently, considerable attention is being paid to proxy-based prefetching because it is more effective and accurate in predicting the correlated pages of many websites of similar interest for more homogeneous users. Compared with client-based prefetching, more complex prediction tasks must run in the proxy, which implies that a more powerful prediction model is required. Thus, based on the time-series characteristics of browsing records, we proposed the intentionality-related long short-term memory (Ir-LSTM) model, which combines both the Skip-Gram embedding method and the LSTM model while expanding the input features with user information. We also propose a novel dynamic allocation module for detecting real-time traffic bursts and correspondingly adjusting the correlation coefficient of the model's output to achieve higher server-side resource utilization while fully maximizing hit ratio.