3 ms·
Don't think bot; think botnet. Ratelimits do not work against botnets since they appear to be independent actors. e.g. if you think it's fine for everyone to
by aidenn0 7y ago
Don't think bot; think botnet. Ratelimits do not work against botnets since they appear to be independent actors. e.g. if you think it's fine for everyone to do something 1-3 times, then you are letting a botnet of 10k hosts do something 10k-30k times.
[edit]
Also, NAT means that there could be hundreds or thousands of individual users on the same IP address (many dorms at smaller colleges are setup this way), so you don't want to rate limit by IP address either.
- dredmorbius 7y agoShouldn't open aggregate listing and blocking of botnets be effective? How many 10k node botnets are there? Also: egress hygiene should be a thing. Block subnets and ASNs if toxic behaviour is detected.
- dmoy 7y ago> How many 10k node botnets are there? About 10 years ago it was somewhere around 10 million bot computers. Obviously the distribution isn't uniform there, but that gives you an idea of the order of magnitude. Also that was 10 years ago, before smart fridges and tvs. So, possibly more now? I don't think "just block all the bots" is a feasible solution here.
- dredmorbius 7y agoNot how many bots, but how many botnets. The nets themselves take resources (time, effort) to set up and maintain. Presumably they're engaged (at least for the purposes of generalised website defence, though specific niches such as targeted commerce fraud may not apply) in systematic behaviour, which leaves signatures. Systemic cross-site collaborative detection. -- essentially what Google's CAPTCHA systems are, though there are others, such as CZ.NIC's Turris project -- could identify these, and via network-based domains of authority (ASN and CIDR assignments) assign reputations and target anti-fraud or anti-abuse countermeasures. Durable, privacy-respecting reputation tokens might be another approach. Present systems are far more primitive and reflect an earlier world-state.
- dheera 7y agoAre there potentially more effective ways to combat this? I'd rather combat malicious behavior than stereotype bots as malicious. That's almost like saying laws don't work against [members of certain race] or [members of certain religion]. Rather we just need some combination of better education and better enforcement strategy instead of stereotyping.