International Journal of Web Engineering and Technology (4 papers in press)
Malicious behavior classification in web logs based on an
improved Xgboost algorithm
by Jiaming Song
Abstract: Attacks against web servers are one of the most serious threats in security fields. Attackers are able to make the computer systems more vulnerable. Analyzing the web logs is one of the most effective methods to identify malicious behaviors. In this study, we consider the analysis of HTTP requests in web logs to classify malicious behavior into multiple categories. At present, web attacks are so complex that single layer classification model is unable to deal with the emerging attacks, in particular, there is a limitation that category features cannot be added to single layer model. Motivated by this, we propose an improved Xgboost algorithm, which uses the method of constructing candidate attacks to attain higher accuracy for malicious behavior detection. The experimental results showthat, compared to other machine learning algorithms, the improved Xgboost algorithm we proposed performs better. Besides, after extracting the important features, it not only does not affect the effectiveness of the algorithm model, but also improves the computational efficiency.
Keywords: web logs; malicious behavior classification; two-layer model; category features; candidate attacks.
Stable web scraping: an approach based on neighbour zone and path similarity of page elements
by Peng Gao
Abstract: Web scraping techniques based on XPath enable users to consistently extract information of interest from webpages that do not provide a structured interface. However, XPath-based extraction is likely to fail when encountering page variants, resulting in a high cost of repair. Countermeasures based on pattern matching or model learning often require careful pre-processing, which is not suitable for cases where the target data is frequently re-designated. In this paper, we present a new extraction method for the stable scraping of arbitrary designated data from webpages. Instead of attempting to find the desired data directly, we first determine its approximate location in the changed page, called the neighbour zone. Then we search for the precise location by ranking the path similarity of page elements within the neighbour zone. Experiments on a large set of real-world webpages show that our method has better stability for web scraping, compared with the XPath-based extraction. In the two datasets, 0.118 and 0.891 F1-score were increased respectively.
Keywords: webpage; web scraping; semi-structured data extraction; XPath expression; stability; HTML tree; node distance; path similarity.
Correlation Based Feature Subset Selection Technique for Web Spam Classification
by Surender Singh
Abstract: In past years different Machine learning Algorithms and Web Spam Features have been created to recognize the Spam. The key part of progression of Machine Learning (ML) depends on the features being utilized. If we have features which correlate with each other then it is easy for ML to learn and if we have features which are very complex then ML may not be able to learn. It is the most imperative and basic area where the majority of the applications in a machine learning are going on. In this paper, Correlation Based Feature Selection (CFS) technique (with Best-First Search) is used which selects features that are most efficient. Two datasets (WebSpam-UK2006 &WebSpam-UK2007) and four classifiers (Na
Keywords: Web spam; machine learning; best first search; correlation-based feature selection.
Sentiment Analysis Based on the Domain Dictionary: A Case of Analsing Online Apparel Reviews
by Ran Tao
Abstract: E-commerce offers an online shopping environment in which manufacturers, businesses, and consumers participate. The past-customer emotion outlined in their reviews plays an important role in not only the purchasing decisions of potential consumers, but also in manufacturers production plans and in business maintenance of their shopping environments. This paper proposes a sentiment analysis approach based on the domain dictionary and a case study as an example to help enterprises and researchers more effectively apply social media analytics in practice. The approach uses web text data acquisition, natural language pre-processing, a domain dictionary, emotion calculation rules, sentiment analysis, and visualization techniques for extracting structured subjective customer emotion and objective commodity metadata from unstructured review pages and analysing the relationships between them. The value of the proposed approach was demonstrated through a case study by using online apparel reviews of an anonymous brand in China in order to understand the relationship between the subjective customer emotion and the objective commodity metadata. The study found that domain dictionaries can be more effective in extracting emotional information. Coats, shirts, and t-shirts garner higher emotions than pants. Sales increase as the level of emotion increases and decrease as the price increases. A fixed price range, in which the clothing has higher emotion and sales, can be found.
Keywords: Data visualisation; Dictionaries; Emotion recognition; Natural language processing; Reviews; Sentiment analysis; User-generated content; Web mining.