基于知识图谱的逻辑验证码方案

Logic CAPTCHA Scheme based on Knowledge Graph

验证码是保护网络资源免受自动化脚本和恶意程序攻击的一种方式。随着人工智能和深度学习技术的不断发展,现阶段利用神经网络对图像进行识别并比较语义标签之间的相似度的方式,能够破解传统的图像验证码方案。因此,提出了一种新的基于知识图谱的逻辑验证码方案。通过在知识图谱ConceptNet中选择常识知识并用词组对其进行描述,使得使用者可以以此进行逻辑推断所描述的内容主体,从而在候选图片中选出合适的答案。它利用人工智能在逻辑推理能力方面的不足,对候选图片添加对抗噪声和对词组生成文本验证码,能有效防止基于图像识别的方法通过判断图片标签是否一致或者语义相似程度对验证码进行破解。实验结果表明,人能够在5 s内以95%的正确率完成该验证码的挑战。因此,逻辑验证码可以成为当前基于文本与图像的验证码的好的替代方案。

CAPTCHA is a way to protect network resources from automated scripts and malicious programs.With the continuous development of artificial intelligence and deep learning technology,the use of neural networks to recognize images and compare the similarity between semantic tags can,at this stage,crack the traditional image CAPTCHA.This paper proposes a new logic CAPTCHA scheme based on the knowledge graph.By selecting common-sense knowledge in knowledge graph ConceptNet and describing it with phrases,the user can logically infer the main content of the description based on this,so as to select a suitable answer from the candidate pictures.It uses the deficiencies of artificial intelligence in logical reasoning ability to add anti-noise to candidate pictures and generate text CAPTCHASs for phrases,which can effectively prevent image recognition methods from cracking CAPCHASs by judging whether the image labels are consistent or semantically similar.Experimental results indicate that people can complete the CAPTCHA challenge with a 95%correct rate in 5 seconds.Therefore logic CAPTCHA can be a good alternative to the current text and image-based CAPTCHA.