DATAKOBOLD CO., LTD. (20240214422). MACHINE LEARNING-BASED HARMFUL-WEBSITE CLASSIFICATION METHOD simplified abstract
Contents
MACHINE LEARNING-BASED HARMFUL-WEBSITE CLASSIFICATION METHOD
Organization Name
Inventor(s)
Nam Goo Song of Suwon-si Gyeonggi-do (KR)
MACHINE LEARNING-BASED HARMFUL-WEBSITE CLASSIFICATION METHOD - A simplified explanation of the abstract
This abstract first appeared for US patent application 20240214422 titled 'MACHINE LEARNING-BASED HARMFUL-WEBSITE CLASSIFICATION METHOD
- Simplified Explanation:**
A machine learning-based method for classifying harmful websites involves tokenizing the HTML source code of a website, vectorizing each token, inputting the vectors into a machine learning model, and determining if the website is harmful.
- Key Features and Innovation:**
- Tokenization and preprocessing of website HTML source code.
- Vectorization of tokens according to a preset algorithm.
- Inputting vector values into a machine learning model for classification.
- Identification of harmful websites based on the model's output.
- Potential Applications:**
This technology can be used in cybersecurity to identify and block harmful websites, protecting users from potential threats such as malware, phishing, and scams.
- Problems Solved:**
The technology addresses the challenge of quickly and accurately identifying harmful websites in real-time, enhancing internet security for users.
- Benefits:**
- Improved detection of harmful websites.
- Enhanced cybersecurity measures.
- Protection against online threats for users.
- Commercial Applications:**
The technology can be utilized by internet service providers, cybersecurity companies, and web browsers to enhance their security features and protect users from malicious websites.
- Questions about Harmful Website Classification:**
1. How does the machine learning model determine if a website is harmful? 2. What are the potential implications of misclassifying a website as harmful or safe?
- Frequently Updated Research:**
Ongoing research in this field focuses on improving the accuracy and efficiency of harmful website classification algorithms, as well as adapting to new types of online threats.
Original Abstract Submitted
a machine learning-based harmful-website classification method performed by a main server includes performing tokenization, by the main server, by extracting and preprocessing an html source code of a website by accessing to the website, vectorizing each token according to a preset algorithm, inputting each vector value is input into a machine learning model, and determining whether the website is a harmful website.