On: Information operations directed at Hong Kong

[bertil]

Flag all the accounts that seem suspicious. The more signal, the easier for Threat Intelligence to identify them.

[29_29 → bertil]

I did, but I feel like this could be solved with machine learning.

[alephnan → 29_29]

Seems like flagging the accounts should, in theory, feed back as data points into their machine learning models?

[jimbokun → 29_29]

That is how supervised machine learning works. Give it know examples, and it extrapolates to identify other instances in new data.

[lucb1e → 29_29]

Flagging bot accounts does not strike me as something a computer would be able to do. If they notice that the machine learned to recognize keywords, they'll use different ones; if they notice it recognizes IP addresses, they'll use different ones; etc. It seems very similar to spam classification to me.