A novel CAPTCHA solver framework using deep skipping Convolutional Neural Networks

Lu, Shida; Huang, Kai; Meraj, Talha; Rauf, Hafiz Tayyab

doi:10.7717/peerj-cs.879

Cited by 7 publications

(3 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A CAPTCHA recognition method with a focal loss function is presented by Wang et al Their method enhances the traditional VGG network structure and incorporates the focal loss function [20]. Lu et al proposed a skip-connection CNN model using two publicly available datasets of text-based CAPTCHA images, which yields a promising result compared to previous studies [21]. More recently, capsule networks have been used due to their capability of preserving detailed information about the input [22].…”

Section: Captcha Recognition With Cnn and Rnnmentioning

confidence: 99%

Adaptive-CAPTCHA: A Text CAPTCHA Solver based on CRNN and Configurable Filter Networks

Wan,

Johari,

Ruslan

2024

Preprint

View full text Add to dashboard Cite

Text-based CAPTCHA remains the most widely adopted security scheme, which is the first barrier to securing websites. Deep learning methods, especially Convolutional Neural Networks (CNNs) are the mainstream approach for text-CAPTCHA recognition, which are widely used in CAPTCHA vulnerability assessment and data collection. However, verification code recognizers are mostly deployed on the CPU platform as part of a web crawler and security assessment, they are required to have both low complexity and high recognition accuracy. Due to the specifically designed anti-attack mechanisms like noise, interference, geometric deformation, twisting, rotation, and character adhesion in text CAPTCHAs, some characters are difficult to efficiently identify with high accuracy in these complex CAPTCHA images. This paper proposed a recognition model named Adaptive-CAPTCHA with a CRNN module and trainable and configurable filtering networks, which effectively handle the interference and learn the correlation between characters in CAPTCHAs to enhance recognition accuracy. Experimental results on two datasets of different complexity show that compared with the baseline model Deep-CAPTCHA, the number of parameters of our proposed model is reduced by about 70% and the recognition accuracy is improved by more than 10 percentage points in the two datasets. In addition, the proposed model has a faster training convergence speed.

show abstract

Section: Captcha Recognition With Cnn and Rnnmentioning

confidence: 99%

Adaptive-CAPTCHA: A Text CAPTCHA Solver based on CRNN and Configurable Filter Networks

Wan,

Johari,

Ruslan

2024

Preprint

View full text Add to dashboard Cite

show abstract

“…A CAPTCHA recognition method with a focal loss function is presented by Wang et al Their method enhances the traditional VGG network structure and incorporates the focal loss function [17]. Lu et al proposed a skip-connection CNN model using two publicly available datasets of text-based CAPTCHA images, which yields a promising result compared to previous studies [18]. A drawback of this technique is the requirement to first segment characters, with the process relying on manually configured operators, which poses a challenge when dealing with CAPTCHAs that contain overlapping characters.…”

Section: Captcha Recognition With Cnn and Rnnmentioning

confidence: 99%

Adaptive CAPTCHA: A CRNN-Based Text CAPTCHA Solver with Adaptive Fusion Filter Networks

Wan,

Johari,

Ruslan

2024

Applied Sciences

View full text Add to dashboard Cite

Text-based CAPTCHAs remain the most widely adopted security scheme, which is the first barrier to securing websites. Deep learning methods, especially Convolutional Neural Networks (CNNs), are the mainstream approach for text CAPTCHA recognition and are widely used in CAPTCHA vulnerability assessment and data collection. However, verification code recognizers are mostly deployed on the CPU platform as part of a web crawler and security assessment; they are required to have both low complexity and high recognition accuracy. Due to the specifically designed anti-attack mechanisms like noise, interference, geometric deformation, twisting, rotation, and character adhesion in text CAPTCHAs, some characters are difficult to efficiently identify with high accuracy in these complex CAPTCHA images. This paper proposed a recognition model named Adaptive CAPTCHA with a CNN combined with an RNN (CRNN) module and trainable Adaptive Fusion Filtering Networks (AFFN), which effectively handle the interference and learn the correlation between characters in CAPTCHAs to enhance recognition accuracy. Experimental results on two datasets of different complexities show that, compared with the baseline model Deep CAPTCHA, the number of parameters of our proposed model is reduced by about 70%, and the recognition accuracy is improved by more than 10 percentage points in the two datasets. In addition, the proposed model has a faster training convergence speed. Compared with several of the latest models, the model proposed by the study also has better comprehensive performance.

show abstract

“…Machine-learning-based schemes and methods have been extensively adopted in different real-time problems. These problems include energy consumption minimization [21], object detection [22], and more specifically the information security domain problems such as credit card fraud detection [23], CAPTCHA solving [24] to enhance the security of CAPTCHA-based security questions. Few of the recent studies focused on feature selection and class imbalance problems while proposing an IDPS.…”

Section: Related Workmentioning

confidence: 99%

An Efficient NIDPS with Improved Salp Swarm Feature Optimization Method

Alabrah

2023

Applied Sciences

View full text Add to dashboard Cite

Network security problems arise these days due to many challenges in cyberspace. The malicious attacks on installed wide networks are rapidly spreading due to their vulnerability. Therefore, the user and system information are at high risk due to network attacks. To protect networks against these attacks, Network Intrusion Detection and Prevention Systems (NIDPS) are installed on them. These NIDPS can detect malicious attacks by monitoring abnormal behavior and patterns in network traffic. These systems were mainly developed using Artificial Intelligence (AI) algorithms. These intelligent NIDPS are also able to detect the attack type while detecting network attacks. Previous studies have proposed many NIDPS for network security. However, many challenges exist so far such as limited available data for training AI algorithms, class imbalance problems, and automated selection of the most important features. These problems need to be solved first, which will lead to the precise detection of network attacks. Therefore, the proposed framework used the highly imbalanced UNSW-NB15 dataset for binary and multiclass classification of network attacks. In this framework, firstly dataset normalization is applied using standard deviation and the mean of feature columns; secondly, an Improved Salp Swarm Algorithm (ISSA) is applied for automated feature selection separately on binary and multiclass subsets. Thirdly, after applying feature selection, the SMOTE–Tomek class balancing method is applied where at least four different ML classifiers are used for binary and multiclass classification. The achieved results outperformed as compared to previous studies and improved the overall performance of NIDPS.

show abstract

A novel CAPTCHA solver framework using deep skipping Convolutional Neural Networks

Cited by 7 publications

References 51 publications

Adaptive-CAPTCHA: A Text CAPTCHA Solver based on CRNN and Configurable Filter Networks

Adaptive-CAPTCHA: A Text CAPTCHA Solver based on CRNN and Configurable Filter Networks

Adaptive CAPTCHA: A CRNN-Based Text CAPTCHA Solver with Adaptive Fusion Filter Networks

An Efficient NIDPS with Improved Salp Swarm Feature Optimization Method

Contact Info

Product

Resources

About