Implement Krum, MultiKrum and FedMedian

[edogab33]

Describe the type of feature and its functionality.

I am working on Federated Learning in adversarial settings, where clients can perform model poisoning in order to degrade the aggregated model performances. To mitigate the effect of model poisoning attacks I have implemented Krum, MultiKrum and FedMedian as strategies.

Would you be interested to include them in the main repository?

Describe step by step what files and adjustments are you planning to include.

I am planning to create three new files in the strategy directory: krum.py, multikrum.py and fedmedian.py.

Is there something else you want to add?

Krum and MultiKrum are described here.

[danieljanes]

@edogab33 sounds great, yes, we'd be interested in adding them!

I'd recomend creating three separate PR's, this will make it easier to review and merge them one by one.

[danieljanes]

Reopening, FedMedian has been merged, but Krum and MultiKrum are still outstanding. We could also created separate issues for them, wdyt @edogab33 ?

[edogab33]

Reopening, FedMedian has been merged, but Krum and MultiKrum are still outstanding. We could also created separate issues for them, wdyt @edogab33 ?

Alright!