,
18 tweets,
5 min read
Read on Twitter
Over the past year, I've been looking at the implications of publicly streaming live webcams in public space and the way these can be matched with social media data to identify ppl in crowds. No special software required. Published today in @BuzzFeedNews
Here's the story. And thanks to the amazing @meghara @HayesBrown and @jsvine
buzzfeednews.com/article/meghar…
buzzfeednews.com/article/meghar…
Check out @meghara the thread (linked above) for an overview and insights into what we found. It was shocking to discover that this was possible - also to realise that anyone with decent resolution footage of a public space could do this - footage is available (or make your own)
It also doesn't rely on expensive or specialist software - while I used common film production tools to do this analysis, simple video viewing software (pre-installed as standard on any consumer laptop) would suffice.
The fact that such high resolution footage was publicly available, is what meant that there was the initial opportunity to experiment that led to the discover that this was possible. The footage is there in the open for anyone to use, not always with the public good in mind
This, for example, was one of the most shocking discoveries to come from this investigation, beyond the initial realisation that that the Instagram sync and subsequent identification was possible - huge amounts of footage have been retained
Also some observations about what made it easy/difficult to geolocate and time stamp videos that might help you think more deeply about how you approach privacy on social media and especially Instagram...
The first and most obvious point, is that all the videos we found and matched were from accounts that were public. It's worth thinking about whether you really need your account to be public. If you're not using insta professionally, maybe you don't?
If you're an artist or photographer using it to show your professional work, it's worth having a public account (and thinking hard about what personal info you divulge in that public account). If it's photos for your friends? Probably best to have it private
In the course of this investigation we found insta bios that contained people's personal phone numbers, for instance (compromising the security of many of their online accounts).
On person filmed the view from a skyscraper. The audio implied it was their first day at work, with a workplace briefing going on in the background. We could identify their new workplace from the visual content of their post. Think about what else you capture as you take a video
Having geolocated and timestamped a video, we could find not only the video maker in webcam footage, but also anyone else in the background of their footage and track them as well, even if identifying them would be slightly harder.
This is worth bearing in mind next time you take and post photos or videos at a protest for example. You are not only compromising your own privacy, bit also that of everyone who appears in the photo or video you post.
A final point about the timing of posts - many people take videos and post them immediately. These were by far the easiest to timestamp and then identify the insta user in thw webcam footage. A delay of even an hour made this harder. Latergram is your privacy friend.
We found the posts in the first place because insta users tagged the location they'd taken the video at - either adding an insta location, or a hashtag. Both can be publicly searched, with top posts collected on a hashtag or location page
Its worth thinking about whether Instagram should have access to your phone's location, necessary for the Instagram location function to work. It is mainly used to allow advertisers and nearby businesses to target you. Turn off location permissions in your settings
And also about the value of adding a hashtag of your location to your post. Again this mainly allows advertisers to target you. It also allows others, from @buzzfeednews to the police, to track your movements, as we showed here.
@BuzzFeedNews Some more thoughts, on the nature of the data that we used here (not what is commonly understood as 'location data') and the worldwide spread of these webcams
