Generative Adversarial Networks in Security: A Survey

Dutta, Indira Kalyan; Ghosh, Bhaskar; Carlson, Albert; Totaro, Michael W.; Bayoumi, Magdy

doi:10.1109/uemcon51285.2020.9298135

Cited by 28 publications

(14 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The usage of a GAN as a defense mechanism renders the model more robust against future attacks. The main objective of a GAN is to detect unknown or unseen attacks and protect systems from various vulnerabilities [18]. In a zero sum game context, a GAN has two NNs competing against each other.…”

Section: Gan-based Defense Methodologymentioning

confidence: 99%

“…Many research papers have employed GANs to either enhance IDS or to design novel attack instances such as generating adversarial malware samples [16,17]. However, there are only a few works on GAN-based intrusion detection [18].…”

Section: Introductionmentioning

confidence: 99%

“…GAN-based adversarial training (AT) can be utilized as a defense mechanism. AT methods inject adversarial examples (AEs) into the training data to make sure that the machine learning technique imparts adversarial perturbations as much as possible [18]. Therefore, they enhance the generalization and resilience of the machine learning techniques, since these models are trained on clean and adversarial examples [19].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Mitigation of Black-Box Attacks on Intrusion Detection Systems-Based ML

et al. 2022

View full text Add to dashboard Cite

Intrusion detection systems (IDS) are a very vital part of network security, as they can be used to protect the network from illegal intrusions and communications. To detect malicious network traffic, several IDS based on machine learning (ML) methods have been developed in the literature. Machine learning models, on the other hand, have recently been proved to be effective, since they are vulnerable to adversarial perturbations, which allows the opponent to crash the system while performing network queries. This motivated us to present a defensive model that uses adversarial training based on generative adversarial networks (GANs) as a defense strategy to offer better protection for the system against adversarial perturbations. The experiment was carried out using random forest as a classifier. In addition, both principal component analysis (PCA) and recursive features elimination (Rfe) techniques were leveraged as a feature selection to diminish the dimensionality of the dataset, and this led to enhancing the performance of the model significantly. The proposal was tested on a realistic and recent public network dataset: CSE-CICIDS2018. The simulation results showed that GAN-based adversarial training enhanced the resilience of the IDS model and mitigated the severity of the black-box attack.

show abstract

Section: Gan-based Defense Methodologymentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Mitigation of Black-Box Attacks on Intrusion Detection Systems-Based ML

et al. 2022

View full text Add to dashboard Cite

show abstract

“…The GAN networks are found to be highly capable of producing plausible results across a wider range of diverse domain applications. The applications such as security [54,55], baggage inspection [56], infected leaf identification [57], covid-19 prediction [58], agriculture [59], business process monitoring [60], Brain MRI synthesis [61], flood visualization [62], estimating the standards of gold [63], ECG wave synthesis [64], Internet of Things (IoT) [65] and Dengue Fever sampling [66].…”

Section: Table 1: Comparison Of Existing Sl Recognition Frameworkmentioning

confidence: 99%

Development of an End-to-End Deep Learning Framework for Sign Language Recognition, Translation, and Video Generation

et al. 2022

View full text Add to dashboard Cite

The recent developments in deep learning techniques evolved to new heights in various domains and applications. The recognition, translation, and video generation of Sign Language (SL) still face huge challenges from the development perspective. Although numerous advancements have been made in earlier approaches, the model performance still lacks recognition accuracy and visual quality. In this paper, we introduce novel approaches for developing the complete framework for handling SL recognition, translation, and production tasks in real-time cases. To achieve higher recognition accuracy, we use the MediaPipe library and a hybrid Convolutional Neural Network + Bi-directional Long Short Term Memory (CNN + Bi-LSTM) model for pose details extraction and text generation. On the other hand, the production of sign gesture videos for given spoken sentences is implemented using a hybrid Neural Machine Translation (NMT) + MediaPipe + Dynamic Generative Adversarial Network (GAN) model. The proposed model addresses the various complexities present in the existing approaches and achieves above 95% classification accuracy. In addition to that, the model performance is tested in various phases of development, and the evaluation metrics show noticeable improvements in our model. The model has been experimented with using different multilingual benchmark sign corpus and produces greater results in terms of recognition accuracy and visual quality. The proposed model has secured a 38.06 average Bilingual Evaluation Understudy (BLEU) score, remarkable human evaluation scores, 3.46 average Fréchet Inception Distance to videos (FID2vid) score, 0.921 average Structural Similarity Index Measure (SSIM) values, 8.4 average Inception Score, 29.73 average Peak Signal-to-Noise Ratio (PSNR) score, 14.06 average Fréchet Inception Distance (FID) score, and an average 0.715 Temporal Consistency Metric (TCM) Score which is evidence of the proposed work.

show abstract

“…GANs have performed well in tasks like image generation and speech synthesis, but have only a few use cases in cybersecurity as surveyed by Dutta et al in [13]. To understand what cybersecurity refers to in the context of this work, the next subsection will explain the background.…”

Section: A Generative Adversarial Networkmentioning

confidence: 99%

Catch Me If You GAN: Using Artificial Intelligence for Fake Log Generation

Toemmel¹

2021

Preprint

View full text Add to dashboard Cite

With artificial intelligence (AI) becoming relevant in various parts of everyday life, other technologies are already widely influenced by the new way of handling large amounts of data. Although widespread already, AI has had only punctual influences on the cybersecurity field specifically. Many techniques and technologies used by cybersecurity experts function through manual labor and barely draw on automation, e.g., logs are often reviewed manually by system admins for potentially malicious keywords.This work evaluates the use of a special type of AI called generative adversarial networks (GANs) for log generation. More precisely, three different generative adversarial networks, SeqGAN, MaliGAN, and CoT, are reviewed in this research regarding their performance, focusing on generating new logs as a means of deceiving system admins for red teams. Although static generators for fake logs have been around for a while, their produces are usually easy to reveal as such. Using AI as an approach to this problem has not been widely researched. Identified challenges consist of formatting, dates and times, and overall consistency. Summing up the results, GANs seem not to be a good fit for generating fake logs. Their capability to detect fake logs, however, might be of use in practical scenarios.

show abstract

Generative Adversarial Networks in Security: A Survey

Cited by 28 publications

References 28 publications

Mitigation of Black-Box Attacks on Intrusion Detection Systems-Based ML

Mitigation of Black-Box Attacks on Intrusion Detection Systems-Based ML

Development of an End-to-End Deep Learning Framework for Sign Language Recognition, Translation, and Video Generation

Catch Me If You GAN: Using Artificial Intelligence for Fake Log Generation

Contact Info

Product

Resources

About