Abstract
With the development of internet communication, spam is quite ubiquitous in our daily life. It not only disturbs users, but also cms. Although there exists many methods of spam detection in both the area of cyber security and natural language processing, their performance is still not capable to satisfy requirements. In this paper, we implemented deep cascade forest for spam detection, a deep model without using backpropagation. With less hyperparameters, the training cost can be easily controlled and declines compared with that in neutral network methods. Furthermore, the proposed deep cascade forest outperforms other machine learning models in the F1 Score of detection. Therefore, considering the lower training cost, it can be considered as a useful online tool for spam detection.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Almeida, T.A.: Sms spam collection data set. https://archive.ics.uci.edu/ml/datasets/SMS+Spam+Collection. Accessed 25 Apr 2019
Almeida, T.A.: SMS spam collection data set from Tiago A. Almeida’s homepage. http://www.dt.fee.unicamp.br/~tiago/smsspamcollection/. Accessed 25 Apr 2019
Almeida, T.A.: Youtube spam collection data set from Tiago A. Almeida’s homepage. http://www.dt.fee.unicamp.br/~tiago//youtubespamcollection/. Accessed 25 Apr 2019
Bengio, Y., Ducharme, R., Vincent, P., Jauvin, C.: A neural probabilistic language model. J. Mach. Learn. Res. 3, 1137–1155 (2003)
Dahl, G.E., Yu, D., Deng, L., Acero, A.: Context-dependent pre-trained deep neural networks for large-vocabulary speech recognition. IEEE Trans. Audio Speech Lang. Process. 20(1), 30–42 (2012)
Dong, C., Zhou, B.: An ensemble learning framework for online web spam detection. In: 2013 12th International Conference on Machine Learning and Applications, vol. 1, pp. 40–45. IEEE (2013)
Drucker, H., Wu, D., Vapnik, V.N.: Support vector machines for spam categorization. IEEE Trans. Neural Netw. 10(5), 1048–1054 (1999)
Hinton, G.E., Salakhutdinov, R.R.: Reducing the dimensionality of data with neural networks. Science 313(5786), 504–507 (2006)
Hochreiter, S., Schmidhuber, J.: Long short-term memory. Neural Comput. 9(8), 1735–1780 (1997)
Jain, G., Sharma, M., Agarwal, B.: Optimizing semantic LSTM for spam detection. Int. J. Inf. Technol. 11(2), 239–250 (2019)
Krizhevsky, A., Sutskever, I., Hinton, G.E.: ImageNet classification with deep convolutional neural networks. In: Advances in Neural Information Processing Systems, pp. 1097–1105 (2012)
Le, Q., Mikolov, T.: Distributed representations of sentences and documents. In: International Conference on Machine Learning, pp. 1188–1196 (2014)
Vergelis, M., Shcherbakova, T., Sidorina, T.: Spam and phishing in 2018. https://securelist.com/spam-and-phishing-in-2018/89701/. Accessed 25 Apr 2019
McCord, M., Chuah, M.: Spam detection on twitter using traditional classifiers. In: Calero, J.M.A., Yang, L.T., Mármol, F.G., García Villalba, L.J., Li, A.X., Wang, Y. (eds.) ATC 2011. LNCS, vol. 6906, pp. 175–186. Springer, Heidelberg (2011). https://doi.org/10.1007/978-3-642-23496-5_13
Ren, Y., Ji, D.: Neural networks for deceptive opinion spam detection: an empirical study. Inf. Sci. 385, 213–224 (2017)
Salton, G.: Developments in automatic text retrieval. Science 253(5023), 974–980 (1991)
Salton, G., Wong, A., Yang, C.S.: A vector space model for automatic indexing. Commun. ACM 18(11), 613–620 (1975)
Sanville, E., Kenny, S.D., Smith, R., Henkelman, G.: Improved grid-based algorithm for bader charge allocation. J. Comput. Chem. 28(5), 899–908 (2007)
Tang, D., Qin, B., Liu, T.: Document modeling with gated recurrent neural network for sentiment classification. In: Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pp. 1422–1432 (2015)
Tracy, M., Jansen, W., Bisker, S.: Guidelines on electronic mail security. NIST Special Publication (2002)
Alberto, T.C., Lochter, J.V., Almeida, T.A.: Youtube spam collection data set. http://archive.ics.uci.edu/ml/datasets/YouTube+Spam+Collection. Accessed 25 Apr 2019
Wu, T., Liu, S., Zhang, J., Xiang, Y.: Twitter spam detection based on deep learning. In: Proceedings of the Australasian Computer Science Week Multiconference, pp. 3:1–3:8. ACM (2017)
Yang, H., Liu, Q., Zhou, S., Luo, Y.: A spam filtering method based on multi-modal fusion. Appl. Sci. 9(6), 1152 (2019)
Zhou, Z.H., Feng, J.: Deep forest: towards an alternative to deep neural networks. In: Proceedings of the 26th International Joint Conference on Artificial Intelligence, pp. 3553–3559. AAAI Press (2017)
Acknowledegment
This work was supported by the National Natural Science Foundation, China (Nos. 61806096, 61872190, 61403208).
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Appendices
A Performance Comparison Between Deep Cascade Forest and Other Classifiers
Tables 3 and 4 shows the precision, recall, precision, accuracy and training/testing time of different models by different kinds of text processing methods on SMS datasets and YouTube datasets, respectively.
B Parameters of LSTM in our Experiment
The LSTM layer contains 64 units for SMS spam detection and 100 units for YouTube spam detection. On both datasets, the batch size is set to 128 In addition, an embedding layer is used to convert each word in the sequence into a dense vector in advance. The embedding layer follows the LSTM layer, a fully connected layer with 256 units, an activation layer using ReLu function, a dropout layer with a dropout rate of 0.1, a fully connected layer with 1 unit, and an activation layer using sigmoid function.
A1. LSTM classification model
The structure of LSTM is shown in Fig. 5, suppose the texts to sequences processing approach is used, then the output integer sequence is fed to the LSTM model. After embedding each token in the sequence into a 50-dimension word vector x, it is then fed to the LSTM layer with 100 hidden units. Overall, the output 100-dimension vector is processed through a 256 units fully connected layer, a 256 units ReLu layer, a fully connected layer with 1 unit in turn, and finally gets the predicted label through the sigmoid mapping.
Rights and permissions
Copyright information
© 2019 Springer Nature Switzerland AG
About this paper
Cite this paper
Chen, K., Zou, X., Chen, X., Wang, H. (2019). An Automated Online Spam Detector Based on Deep Cascade Forest. In: Liu, F., Xu, J., Xu, S., Yung, M. (eds) Science of Cyber Security. SciSec 2019. Lecture Notes in Computer Science(), vol 11933. Springer, Cham. https://doi.org/10.1007/978-3-030-34637-9_3
Download citation
DOI: https://doi.org/10.1007/978-3-030-34637-9_3
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-34636-2
Online ISBN: 978-3-030-34637-9
eBook Packages: Computer ScienceComputer Science (R0)