Deep Learning for Anomaly Detection : A Survey

The aim of this survey is two fold, firstly we present a structured and comprehensive reviewof research methods in deep anomaly detection (DAD). Furthermore, we also discuss the adoption of DAD methods across various application domains and assess their effectiveness.

Contributing

Please leave a comment and create a new issue @ Include My Work. if you wish your work be included in the survey. Kindly request you to provide us the title of paper and preferably bibtex entry.

I would be updating the paper every two-weeks and will be including and citing your paper in the appropriate sections of research or application sections. This enables us to keep track and follow the recent progress in the space of deep learning based techniques for anomaly detection.

Survey paper can be downloaded from Arxiv.

Table of contents

1. Introduction

   1. What are anomalies?
   2. What are novelties?
   3. Motivation and Challenges: Deep anomaly detection (DAD) techniques
   4. Related Work
   5. Our Contributions
   6. Organization

2. Different aspects of deep learning-based anomaly detection.

   1. Nature of Input Data
   2. Based on Availability of labels
      • Supervised deep anomaly detection
      • Semi-supervised deep anomaly detection
      • Unsupervised deep anomaly detection
   3. Based on training objective
      • Deep Hybrid Models(DHM)
      • One-Class Neural Networks (OC-NN)
   4. Type of Anomaly
      • PointAnomalies
      • Contextual Anomaly Detection
      • Collective or Group Anomaly Detection
   5. Output of DAD Techniques
      • AnomalyScore
      • Labels

3. Applications of Deep Anomaly Detection

   1. Intrusion Detection
   2. Host-Based Intrusion Detection Systems (HIDS)
   3. Network Intrusion Detection Systems (NIDS)
   4. Fraud Detection
   5. Banking fraud
   6. Mobile cellular network fraud
   7. Insurance fraud
   8. Healthcare fraud
   9. Malware Detection
   10. Medical Anomaly Detection
   11. Deep learning for Anomaly detection in Social Networks
   12. Log Anomaly Detection
   13. Internet of things (IoT) Big Data Anomaly Detection
   14. Industrial Anomalies Detection
   15. Anomaly Detection in TimeSeries
   16. Video Surveillance 4.#### Deep Anomaly Detection(DAD) Models
   17. Supervised deep anomaly detection
   18. Semi-supervised deep anomaly detection
   19. Hybriddeepanomalydetection
   20. One-class neural networks (OC-NN) for anomaly detection
   21. Un-supervised Deep Anomaly Detection
   22. Miscellaneous Techniques
       • Transfer Learning based anomaly detection
       • Zero Shot learning based anomaly detection
       • Ensemble based anomaly detection
       • Clustering based anomaly detection
       • Deep Reinforcement Learning (DRL) based anomaly detection
       • Statistical techniques deep anomaly detection

4. Deep neural network architectures for locating anomalies

   1. Deep Neural Networks (DNN)
   2. Spatio Temporal Networks(STN)
   3. Sum-Product Networks(SPN)
   4. Word2vec Models
   5. Generative Models
   6. Convolutional Neural Networks
   7. Sequence Models
   8. Autoencoders

5. Relative Strengths and Weakness : Deep Anomaly Detection Methods

6. Conclusion

Authors

• Raghavendra Chalapathy - Initial work

See also the contributors referenced within the paper .

License

This project is licensed under the MIT License - see the LICENSE.md file for details

Citation and Contact

You find the PDF of the Deep Learning for Anomaly Detection : A Survey paper at

If you use our work, please also cite the paper:

If you have any suggestions about paper, feel free to mail me :)