An explainable and efficient deep learning framework for video anomaly detection

Wu, Chongke; Shao, Sicong; Tunc, Cihan; Satam, Pratik; Hariri, Salim

doi:10.1007/s10586-021-03439-5

Cited by 34 publications

(8 citation statements)

References 54 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Ranganathan et al [9] have used the Diffie-Hellman key exchange technique and the advanced encryption standards (AES) technique to implement the concept of differential privacy, which are both quite powerful in terms of speed. e tests were conducted with Laplace and Gaussian methods, which are the techniques currently most commonly used.…”

Section: Related Workmentioning

confidence: 99%

“…Breaches are unavoidable, threats have become more complex, and database security has become more difficult. Furthermore, many threats are undetectable by traditional policy-based or rule-based security systems [9]. Firewalls, access control levels, and rule-based management are useless in circumstances of stolen privileged user accounts or internal attacks.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Method of Cumulative Anomaly Identification for Security Database Based on Discrete Markov chain

Yang

Najafi

2022

Security and Communication Networks

View full text Add to dashboard Cite

There exists an enormous volume of data in the database system, which is accountable for the storage of data and organization of data. The intruders can breach the security system of database and steal the important information. Therefore, it is of great significance to carry out the cumulative anomaly identification of the security database. In view of the shortcomings of traditional anomaly detection methods in detection performance and poor effect of anomaly recognition, this paper proposes a cumulative anomaly recognition method based on discrete Markov chain for security database. First, the sniffer is used to read the user access behaviour data, and then, it is processed, that is, standardized processing. Then, the segmentation method is used to extract the user behaviour features, and the normal feature data and abnormal feature data are obtained. Finally, the state sequence generated by the discrete Markov chain is used to calculate the state probability, which is used to evaluate the abnormal process behaviour. The proposed method in this paper is based on the Markov chain and can be used for better anomaly recognition. The results are obtained in terms of sensitivity score, precision score, and F1-score. The results are also compared with the results obtained by using some of the state-of-the-art traditional techniques. The comparison clearly indicated that the proposed method is more effective as compared to the tradition methods. The proposed method has the highest F1-score of 0.8586, and then the traditional methods have F1-scores of 0.7233, 0.8236, and 0.7562 for methods 1, 2, and 3, respectively.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Method of Cumulative Anomaly Identification for Security Database Based on Discrete Markov chain

Yang

Najafi

2022

Security and Communication Networks

View full text Add to dashboard Cite

show abstract

“…A new cloud data platform login and identity verification system was proposed, which is equipped with a fingerprint authentication and identification module, which has the advantages of high information security and strong antiattack ability [17]. After testing, this technology improves the security and validity of the connection and access to the cloud data platform without affecting the user experience [18]. e security performance requirements of the cloud data platform application in the communication process authentication were analyzed.…”

Section: Related Workmentioning

confidence: 99%

Cloud Computing Database and Travel Smart Platform Design Based on LSTM Algorithm

Chen

2022

Mobile Information Systems

View full text Add to dashboard Cite

Information technology has played a key role in the development of the tourism planning service industry and has now become an important foundation for the survival and rapid development of the industry. In this context, due to the fast updating and popularization of information technology, it has greatly promoted the development of the tourism industry. In order to meet the current public demand for tourism information, this paper integrates cloud computing, VR technology, and big data analysis technology to build a smart platform for intelligent perception tourism information services. The system can obtain tourism information through mobile Internet terminals. Among them, the database of the smart tourism planning platform is the most important module. Aiming at the many difficulties in adapting data encryption in cloud storage applications, this article designs an adaptive CloudCrypt data encryption system based on cloud computing technology and proposes dynamic JavaScript dynamic analysis and automatic identification of data technology, through adaptive different cloud applications to obtain data encryption protection. CloudCrypt is suitable for typical cloud applications, such as mail and storage. The entry cost of the system is extremely low; it can fully guarantee the security of the tourism information platform database and can integrate wireless sensor networks into the tourism information platform. The network system composed of sensor nodes activates detection, calculation, and communication modules through wireless communication and has the advantages of low cost, low power consumption, and fast networking speed. In this paper, through the construction of integrated wireless sensor technology and cloud computing database, it is applied to the construction of a tourism smart platform, thereby promoting the development of smart travel technology.

show abstract

“…(AUC) MDT [7] 2010 82.9% N/A Conv-AE [8] 2015 85.0% 80.0% ConvLSTM-AE [11] 2017 88.1% 77.0% Unmasking [12] 2017 82.2% 80.6% Stacked RNN [18] 2017 92.2% 81.7% MCNet [14] 2017 95.0% 83.8% Liu et al [15] 2018 95.4% 85.1% RC-GAN [17] 2019 93.5% 82.1% MemAE [13] 2019 94.1% 83.3% Hu et al [30] 2020 90.4% / DAE-SHAP [33] 2021 92.4% / F2PN [19] 2021 96.2% 85.7% Sun et al [31] 2022 / 85.7% DAGMM [32] 2022 95.7% / Wang et al [34] 2022 95.6% 73.1% NA-UnetG [20] 2022 96.3% 83.4% Ours 2022 96.9% 86.1%…”

Section: Cuhk Avenuementioning

confidence: 99%

Video Anomaly Detection Based on Optical Flow Feature Enhanced Spatio–Temporal Feature Network FusionNet-LSTM-G

et al. 2022

View full text Add to dashboard Cite

Video anomaly detection refers to the rapid identification and localization of anomalous events during video surveillance. It has become a challenging task due to the dependence of anomalous event definitions on specific scenes and the scarcity, diversity and non-exhaustiveness of anomalous events. This manuscript proposes a robust frame prediction-based method for video anomaly detection, it can solve the problems of underutilized spatio-temporal sequence information and difficulty in capturing subtle motions. Firstly, a spatial-temporal feature fusion network is constructed using FusionNet and LSTM. Secondly, feature optimization is performed on this spatio-temporal feature fusion network using optical flow feature enhancement. To further improve the robustness of the model, patch GAN is added to FusionNet-LSTM for adversarial training, and the objective function in training is optimized with apparent constraints and image similarity constraints to approach the real frame. Finally, to reduce the sequence error, a PNSR-based mean normalized anomaly score calculation is used in the manuscript. Experiments show that the average accuracy of the proposed method for anomalous event detection on UCSD Ped2 and CUHK Avenue datasets is 96.9% and 86.1% respectively, which is 1.5% and 1.0% better than the benchmark model U-Net，and the model has low computational complexity where FLOP is 6.196G and the number of parameters is 6.829M. It can meet the application requirements of real-time video anomaly detection and has great value for security video surveillance.

show abstract

An explainable and efficient deep learning framework for video anomaly detection

Cited by 34 publications

References 54 publications

Method of Cumulative Anomaly Identification for Security Database Based on Discrete Markov chain

Method of Cumulative Anomaly Identification for Security Database Based on Discrete Markov chain

Cloud Computing Database and Travel Smart Platform Design Based on LSTM Algorithm

Video Anomaly Detection Based on Optical Flow Feature Enhanced Spatio–Temporal Feature Network FusionNet-LSTM-G

Contact Info

Product

Resources

About