Couple of Questions

[Tim-Cunnane]

Hi Folks,

I just have a couple of question to ask regarding the system and hopefully you will be able to let me know of the answers:

In terms of other pcaps that could be used for the system, would you have any other pcaps that could be used for the system such as DDoS, Inflitration or any other pcaps, to allow for models to be built?

For the model comparison score what how would this be calculated and what is this indicating?

Thanks
Tim

[strongcourage]

@Tim-Cunnane: This issue arises because the tool currently supports executing one action at a time, such as building a model, to prevent potential conflicts. However, it can be easily modified to support multiple actions simultaneously.

As for datasets, we used the public dataset available at this link. Regarding model comparison, our current approach involves using the confusion matrix for different cutoff values. However, there are various other metrics available for different analytical purposes.

[Tim-Cunnane]

@strongcourage ahhh right okay thanks for help on that. For the pcaps, I extract some pcap for DDoS etc. but was wondering do you need to extract normal pcap for each attack and how you obtained it?

[strongcourage]

@Tim-Cunnane: From this paper (Section 4 - Dataset), they captured both normal traffic on Monday and malicious traffic on the rest of the week. Thus, you can obtain both benign and malicious pcaps.

[Tim-Cunnane]

@strongcourage perfect hopefully be able to get some pcaps from this.

I don’t know if there is any, but would there be any testing for maip like CI testing, testing of the apis, etc.?

[strongcourage]

@Tim-Cunnane: the project is still under development as we're working on and testing some functionalities.
Thank you for your suggestions, for now only manual testing :D. Indeed implementing CI testing would be a beneficial approach, I will definitely consider it.

[Tim-Cunnane]

@strongcourage In terms of the normal pcaps that have to enter into the build of a model, do normal pcaps and malicious pcap to all for the model to be built. I have create a few normal pcaps however the don't seem to provide any performance metrics. I have a range of attacks like DDoS, Infiltration and brute force pcap to test against the system.

[strongcourage]

@Tim-Cunnane: if possible, can you share with me normal/malicious pcap files you've used for building the model? I will try to take a look on them. Thanks.

[Tim-Cunnane]

@strongcourage Here is a link to some of the pcaps that I have currently.

https://qubstudentcloud-my.sharepoint.com/:f:/g/personal/40294818_ads_qub_ac_uk/EhoVqdcTcDFAnwKR4XgN48wBU_fcxxtQZ2xh5kxsPT8o-w?e=TzkPPX

Thanks

[Tim-Cunnane]

@strongcourage have you been able to take a look those pcaps? Apologies for the rush, just have this final section of my dissertation to complete.

Thanks
Tim

[Tim-Cunnane]

@strongcourage When perform the GAN-driven attack how long does this usually take? Does this create the synthetic data?
Currently getting this issue when perform this attack:

Thanks
Tim

[Tim-Cunnane]

@strongcourage would there be any advise that you could provide that would fix this issue?

Thanks

[strongcourage]

@Tim-Cunnane: Hello, as I mentioned in this comment, the issue here is that you don't have ctgan_samples.csv in the folder /maip/src/server/deep-learning/attacks/model_14%20(DDoS). You need to generate this dataset using CTGAN, or just copy ctgan_samples.csv from other folders (I'm not sure it is correct for your new model). Hope this helps.

[Tim-Cunnane]

@strongcourage: Hello, using the system with docker containers, when a new model is created it doesn’t create a new folder within models same happens with the attacks.

Thanks
Tim

[Tim-Cunnane]

@strongcourage, Hi, this issue still seem to occur when the samples are in here as seen in the below:


Would there be any specific reason for this when the platform is ran in an containerised format?
Thanks

[strongcourage]

@Tim-Cunnane: maybe you can change your model to model_14_ddos then try again. Anyway, ctgan_poisoned_dataset.csv is not successfully generated.