I've been wanting to do some social graph experimentation on it (small scale - say 1000 people near me) but concluded I probably couldn't scrape enough via raw scraping without freaking out their anti-scraping. (And API is a non-starter since that basically says everything is verboten).
Here are some tricks which may or may not work today:
- Have an app where user logs in through said website, then scrape their friends using this user's token. That way you get exponential leverage on the number of API calls you can make, with just a handful of users.
- Call their API through ipv6, because they may not yet have a proper, ipv6 subnet-based rate limiter.
- Scrape the mobile website. Even Facebook still has a non-js mobile version. This single WAP/mobile website defeats every anti-scraping measure they may have.
- From a purely practical perspective, start with a baremetal transaction-isolation-less database like Cassandra/ScyllaDB. Don't rely on googling "postgres vs mongodb" or "sql vs nosql", those articles will all end in "YMMV". What you really need is massive IOPS, and a multi-node ring-based index with ScyllaDB will achieve that easily. Or just use MongoDB on one machine if you're not in hurry.
- Don't be too kind on the big websites. They can afford to keep all their data in hot pages, and as a one man you will never exhaust them.
Nice tip!!
> -- From a purely practical perspective, start with a baremetal transaction-isolation-less database like Cassandra/ScyllaDB. Don't rely on googling "postgres vs mongodb" or "sql vs nosql", those articles will all end in "YMMV". What you really need is massive IOPS, and a multi-node ring-based index with ScyllaDB will achieve that easily. Or just use MongoDB on one machine if you're not in hurry.
Somewhat ironically Elasticsearch would probably work really well for this too (just make sure your elasticsearch isn't open to the world on the internet!).
Sure it will work, but I personally don't like Elasticsearch for anything high-intensity because of its HTTP REST API and the overhead it carries. Take a look at Cassandra's [1] "CQL binary protocol", it simple and always on point.
[1] https://github.com/apache/cassandra/blob/trunk/doc/native_pr...
In all seriousness does anyone know why you can even host an elasticsearch database as http and without credentials? Seems to be the default. What is the use case for this?
For a while I've had reoccurring nightmares that my DB had been stolen and published together with an article on how stupid and incompetent I am.
I'll cut straight to the chase and post it on hn. This intermediate step of waiting for someone to discover it takes too long
That's some extremely shady thing to do.
I usually recommend latency-based dynamic load control for that. Once the website starts to reply 500-1000ms longer than the average one-thread latency, it is time to take a bit of it back. It is also a co-operative strategy between fellow scrapers, even if they don't know about the other ones pushing larger load on the servers.
YMMV, and cloud providers would hate you for this, but you can automate the IP rotation with a cloud providers that bills you by the hour. It's easier than ever nowadays to spin an instance in Frankfurt, use it for an hour, and then another in Singapore for the second hour.
Pretending to be Googlebot also helps.
Clever. VMs with IPV6 are cheap as a bonus :)
Same for non-js mobile. Thanks for the tips
How would someone do that using node.js? Asking for a friend.
A more useful answer is: I did this once, many years ago. Back then it was a matter of hooking up PhantomJS and making sure your user string was set correctly. Since PhantomJS was – I think – essentially the same as what headless chrome is today, the server can't determine that you're running a headless browser.
Now, it's not so easy nowadays to do that. There are mechanisms to detect whether the client is in headless mode. But most websites don't implement advanced detection and countermeasures. And in the ideal case, you can't really detect that someone is doing automated scraping. Imagine a VM that's literally running chrome, and the script is set up to interact with the VM using nothing but mouse movements and keyboard presses. You could even throw in some AI to the mix: record some real mouse movements and keyboard presses over time, then hook up some AI to your script such that it generates movements and keyboard presses that are impossible to distinguish from real human inputs. Such a system would be almost impossible to differentiate vs your real users.
The other piece of the puzzle is user accounts. You often have to have "aged" user accounts. For example, if you tried to scrape LinkedIn using your own account, it wouldn't matter if you were using 500 IPs. They would probably notice.
It's hard to counter a determined scraper.
It's not very hard to get something that would be too hard for almost every website beside Google and Facebook to bother with. If it's a 1 on a 0-9 scale in difficulty, most websites just don't have the resources to detect it
It took me like ~3 hours to write it, but I guarantee it would take months for someone to detect it, and even then, they'd have a lot of false positives and negatives.
How many fines has GDPR resulted in?
Headless chrome cat and mouse game is a lot of fun. We need more players.
I had a business that was generating more money than my full-time job for a while. We helped and greatly simplified matters for several thousand independent proprietors while having a positive effect on the load of the data source, since we were able to batch/coalesce requests, make better use of caches, and take notification responsibilities on ourselves.
Once in a while someone would get worried and grumpy at the data source and there were a couple of cat-and-mouse games, but we easily outwitted their scraping detection each time. When they got tired of losing the technical game, they sent out the lawyers, which was far more effective. We were acquiring facts about dates and times from the place that issued/decided those dates and times, so there wasn't really any reliable alternative data source, and we had to shut down.
The glimmer of hope on the horizon is LinkedIn v. HiQ, which seems poised to potentially finally overturn 4 decades of anti-scraping case law, but not holding my breath too hard there.
> In a long-awaited decision in hiQ Labs, Inc. v. LinkedIn Corp., the Ninth Circuit Court of Appeals ruled that automated scraping of publicly accessible data likely does not violate the Computer Fraud and Abuse Act (CFAA). This is an important clarification of the CFAA’s scope, which should provide some relief to the wide variety of researchers, journalists, and companies who have had reason to fear cease and desist letters threatening liability simply for accessing publicly available information in a way that publishers object to. It’s a major win for research and innovation, which will hopefully pave the way for courts and Congress to further curb abuse of the CFAA.
https://www.eff.org/deeplinks/2019/09/victory-ruling-hiq-v-l...
In either case my personal data is given away without my consent, but there's this implication that it's only an issue when someone doesn't pay for it.
Normally the company sells this data, but now they've given it away. It's not good this data got out because the curation has some value to spammers or whoever. But using the word "leak" here undermines the severity of a real leak where passwords and social security numbers are exposed. Data that was never meant by anyone to be open.
Everyone likely has (technically) provided consent for every piece of information here being shared with partners. Buried in fine print that it wasn't really expected they'd read, of course. It's the cost of being online, and that sucks, but it seems only a leak of what had already been given out.
You gave that consent when you put your info in Linkedin in the first place, according to their ToS.
[0] https://www.abcactionnews.com/news/local-news/i-team-investi...
Mind you, I didn't research the topic of what can or cannot be requested with FOIA, so I might be totally wrong.
Needing to be logged in as the same user defeats the purpose of proxying to hide your physical origin.
Registering thousands of different users to use in a distributed way is hard now that they require a text message verification for new accounts.
https://www.eff.org/deeplinks/2019/09/victory-ruling-hiq-v-l...
They were ordered to unblock hiQ specifically, they were not ordered to open up content to scrapers generally.
They can still throttle high volume traffic and put up captchas. I think the only specific thing the court ordered was for them to unblock hiQ IP ranges.
This may or may not still work.