ICACT20210214 Question.1
Questioner: shahram@ieee.org 2021-02-09 오후 2:44:50 |
ICACT20210214 Answer.1
Answer by Auhor baniyab@gram.edu 2021-02-09 오후 2:44:50
Chrome  Click!! |
| Given that your feature selection works based on the existing training dataset, how would your proposed method perform if it is tested on an atypical attack (with new features), or a polymorphic attack that changes its features dynamically? |
Thank you for your genuine question Dr. Shahram, my understanding is that the attack pattern (polymorphic attack) somehow distinct (initially it may like normal when time progresses attack pattern deviated from normal traffic). Even, it may be different from than originally trained attack pattern. The trained model certainly recognized the attack (because it already learned the probably normal pattern too). However, it is hard to tell the exact prediction rate quantitatively (without experiment). |
ICACT20210214 Question.2
Questioner: wkdgus4788@chungbuk.ac.kr 2021-02-09 오후 2:46:16 |
ICACT20210214 Answer.2
Answer by Auhor baniyab@gram.edu 2021-02-09 오후 2:46:16
Chrome Click!! |
| By any chance, do you have any plans to test it as other datasets like UNSW-NB15? |
Yes, I am planning to test UNSW-NB15 and other datasets too. Thank you for your question.
|
ICACT20210214 Question.3
Questioner: rashid_zaidi@yahoo.com 2021-02-11 오전 1:35:59 |
ICACT20210214 Answer.3
Answer by Auhor baniyab@gram.edu 2021-02-11 오전 1:35:59
Chrome Click!! |
| It is very interesting idea. You think there is more room of improvement in you IDS algorithm? |
Thank you for your question, Rashid, definitively, we are a lot of room for improvement. Now, we extending this trained model for detecting the polymorphic (dynamic features) malware using LSTM and deep learning algorithms. We also plan to use other feature selection algorithms for comparison. |
