Spam Image Detection Model based on Deep Learning for Improving Spam Filter
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Due to the development and dissemination of modern technology, anyone can easily communicate usingservices such as social network service (SNS) through a personal computer (PC) or smartphone. The developmentof these technologies has caused many beneficial effects. At the same time, bad effects also occurred, oneof which was the spam problem. Spam refers to unwanted or rejected information received by unspecifiedusers. The continuous exposure of such information to service users creates inconvenience in the user's use ofthe service, and if filtering is not performed correctly, the quality of service deteriorates. Recently, spammersare creating more malicious spam by distorting the image of spam text so that optical character recognition(OCR)-based spam filters cannot easily detect it. Fortunately, the level of transformation of image spamcirculated on social media is not serious yet. However, in the mail system, spammers (the person who sendsspam) showed various modifications to the spam image for neutralizing OCR, and therefore, the same situationcan happen with spam images on social media. Spammers have been shown to interfere with OCR readingthrough geometric transformations such as image distortion, noise addition, and blurring. Various techniqueshave been studied to filter image spam, but at the same time, methods of interfering with image spamidentification using obfuscated images are also continuously developing. In this paper, we propose a deeplearning-based spam image detection model to improve the existing OCR-based spam image detectionperformance and compensate for vulnerabilities. The proposed model extracts text features and image featuresfrom the image using four sub-models. First, the OCR-based text model extracts the text-related features,whether the image contains spam words, and the word embedding vector from the input image. Then, theconvolution neural network-based image model extracts image obfuscation and image feature vectors from theinput image. The extracted feature is determined whether it is a spam image by the final spam image classifier. As a result of evaluating the F1-score of the proposed model, the performance was about 14 points higher thanthe OCR-based spam image detection performance.
Abstract
Due to the development and dissemination of modern technology, anyone can easily communicate usingservices such as social network service (SNS) through a personal computer (PC) or smartphone. The developmentof these technologies has caused many beneficial effects. At the same time, bad effects also occurred, oneof which was the spam problem. Spam refers to unwanted or rejected information received by unspecifiedusers. The continuous exposure of such information to service users creates inconvenience in the user's use ofthe service, and if filtering is not performed correctly, the quality of service deteriorates. Recently, spammersare creating more malicious spam by distorting the image of spam text so that optical character recognition(OCR)-based spam filters cannot easily detect it. Fortunately, the level of transformation of image spamcirculated on social media is not serious yet. However, in the mail system, spammers (the person who sendsspam) showed various modifications to the spam image for neutralizing OCR, and therefore, the same situationcan happen with spam images on social media. Spammers have been shown to interfere with OCR readingthrough geometric transformations such as image distortion, noise addition, and blurring. Various techniqueshave been studied to filter image spam, but at the same time, methods of interfering with image spamidentification using obfuscated images are also continuously developing. In this paper, we propose a deeplearning-based spam image detection model to improve the existing OCR-based spam image detectionperformance and compensate for vulnerabilities. The proposed model extracts text features and image featuresfrom the image using four sub-models. First, the OCR-based text model extracts the text-related features,whether the image contains spam words, and the word embedding vector from the input image. Then, theconvolution neural network-based image model extracts image obfuscation and image feature vectors from theinput image. The extracted feature is determined whether it is a spam image by the final spam image classifier. As a result of evaluating the F1-score of the proposed model, the performance was about 14 points higher thanthe OCR-based spam image detection performance.
참고문헌 (24)
- Mozhgan Malekshahi Rad, "Social Internet of Things: vision, challenges, and trends", Human-centric Computing and Information Sciences, 10(1) (2020), Springer Science and Business Media LLC[DOI]
- Joni Salminen, "Developing an online hate classifier for multiple social media platforms", Human-centric Computing and Information Sciences, 10(1) (2020), Springer Science and Business Media LLC[DOI]
- Z. Zhang, "A crowdsourcing method for online social networks security assessment based on human-centric computing", Human-centric Computing and Information Sciences, 10 (2020)
- Jin Ho Park, "A Comprehensive Survey on Core Technologies and Services for 5G Security: Taxonomies, Issues and Solutions", Human-centric Computing and Information Sciences, 1(0), pp.1~22 (2021), KIPS-CSWRG[DOI]
- S. Rathore, "Deep learning and blockchain-empowered security framework for intelligent 5G-enabled IoT", IEEE Access, 9, pp.90075~90083 (2021)
- S. Rathore, "A blockchain-based deep learning approach for cyber security in next generation industrial cyber-physical systems", IEEE Transactions on Industrial Informatics, 17(8), pp.5522~5532 (2021)
- "Apache SpamAssassin"[URL]
- B. Biggio, "Image spam filtering using visual information", Proceedings of the 14th International Conference on Image Analysis and Processing (ICIAP), pp.105~110 (2007)
- Z. Wang, "Filtering image spam with near-duplicate detection", Proceedings of the 4th Conference on Email and Anti-Spam (CEAS) (2007)
- A. Barbar, "Artificial Intelligence and Applied Mathematics in Engineering Problems", pp.347~364 (2020), Springer
- C. Fatichah, "Intelligent and Interactive Computing", pp.295~303 (2019), Springer
- N. Imam, "Detecting spam images with embedded Arabic text in twitter", Proceedings of 2019 International Conference on Document Analysis and Recognition Workshops (ICDARW), pp.1~6 (2019)
- A. P. Singh, "Image spam classification using deep learning", Master, pp.2018
- S. Rao, "Computational Vision and Bio-Inspired Computing", pp.1378~1385 (2020), Springer
- A. Fan, "Image spam filtering using convolutional neural networks", Personal and Ubiquitous Computing, 22, pp.1029~1037 (2018)
- T. Sharmin, "Convolutional neural networks for image spam detection", Information Security Journal: A Global Perspective, 29(3), pp.103~117 (2020)
- "TorchText"[URL]
- "Affine Transformations tutorials" (2019)[URL]
- "Smoothing images tutorial: Gaussian Blurring" (2019)[URL]
- "CAPTCHA: Telling Humans and Computers Apart Automatically" (2010)[URL]
- G. Bradski, "The OpenCV library", Dr. Dobb's Journal, 25(11), pp.120~125 (2000)
- "Image Spam Dataset" (2007)[URL]
- "Unsplash"[URL]
- "Tesseract-OCR" (2023)[URL]