**3. Conclusion**

This development also explored the efficacy of the deep learning approach, which is part of the set objective to explore relevant algorithm for the detection of phishing, this revealing the advantages and disadvantages of both the convolutional neural network (CNN) and long short-term memory (LSTM) methods. On the one

hand, the LSTM+CNN algorithm was also used to develop an offline approach for phishing detection but had a smaller detection accuracy of 93.28% compared to that of the ANFIS algorithm.

The reduction in the number of features makes this much faster in terms of time-to-prediction. The protection aspect of the solution is implemented via a user warning interface with various colours representing the category of detection. A green colour indicates a legitimate site, whilst an amber colour represents suspicious ones, and a red colour indicates a phishing site. There is also an audible (voice) warning of relevance to a visually impaired person. The protection interface also advises the user on what to do next such as to terminate the process if it discovers that the site is phishing or suspicious.

The development reflects the effectiveness of the hybrid features approach using CNN, and the LSTM deep learning algorithm is an essential driver to the high model performance. This chapter has contributed to the anti-phishing detection research by present the use of a hybrid feature which include image, frame and text. These three sets of input have just been introduced as single hybrid features for the first time. The three elements are used because they represent the whole structure of a website. Although the scheme performed well, parameter tuning influenced the algorithm in a positive way, and it must be pre-specified to solve a given problem. Ultimately online user confidence will increase in performing transactions online.

The main conclusion of applying the IPDSS approach that is in this development achievement an excellent classification accuracy of 93.28% for identifying phishing websites.
