WEBVTT - Somebody's Watching All of Us 0:00:03.600 --> 0:00:06.640 How do you plan to spend the holidays? Don't answer 0:00:07.040 --> 0:00:10.080 the company's safe Graph probably has a good idea already. 0:00:10.240 --> 0:00:17.240 I'm Jonathan Strickland and this is text Uff Daily. Adrian 0:00:17.400 --> 0:00:19.919 Jeffreys wrote a piece for the Outline dot com that 0:00:20.079 --> 0:00:23.239 was more than a little alarming. Jeffreys reported on a 0:00:23.280 --> 0:00:26.600 study that found family members who subscribed to different political 0:00:26.640 --> 0:00:31.600 ideologies spend less time together during holidays. That in itself 0:00:31.720 --> 0:00:35.320 isn't alarming. What was alarming was the source of the 0:00:35.320 --> 0:00:39.440 researchers data from which they drew this conclusion. That source 0:00:39.560 --> 0:00:42.800 was a company called safe Graph, which gave the researchers 0:00:42.840 --> 0:00:47.520 access to seventeen trillion location markers that came from ten 0:00:47.680 --> 0:00:51.880 million smartphones. That means the researchers had seventeen trillion points 0:00:51.920 --> 0:00:54.880 of data related to those smartphones locations over the course 0:00:54.920 --> 0:00:59.440 of the sixteen Thanksgiving holiday. Even your most dedicated gum 0:00:59.480 --> 0:01:02.240 shoe isn't likely to rack up that much information about 0:01:02.280 --> 0:01:05.880 to mark their tailing. Using this data, the researchers were 0:01:05.880 --> 0:01:09.559 able to extrapolate some other information. For example, they worked 0:01:09.600 --> 0:01:12.600 under the assumption that if a smartphones location was frequently 0:01:12.640 --> 0:01:15.039 in the same spot between the hours of one and 0:01:15.160 --> 0:01:18.399 four am, it was most likely the home of the 0:01:18.440 --> 0:01:21.880 smartphone's owner. Sure, you might go out clubbing once in 0:01:21.920 --> 0:01:24.039 a while, but probably not every day of the week 0:01:24.560 --> 0:01:28.120 unless you're Steffon. They also looked at location data for 0:01:28.160 --> 0:01:31.080 mobile devices during the hours of one and five PM 0:01:31.160 --> 0:01:33.839 on Thanksgiving Day to see how many people were spending 0:01:33.840 --> 0:01:37.240 the holiday at a different location or in transit. On 0:01:37.280 --> 0:01:40.920 the surface, this data is anonymous. There's no overt link 0:01:41.000 --> 0:01:45.080 between the location data and any individual's identity. However, it 0:01:45.120 --> 0:01:48.480 doesn't take much sleuthing to figure out which smartphone belongs 0:01:48.520 --> 0:01:52.160 to any particular person. If I look at detailed location 0:01:52.200 --> 0:01:54.760 markers linked to your device, and I have a passing 0:01:54.800 --> 0:01:57.800 familiarity with your habits, it's probably not going to take 0:01:57.840 --> 0:01:59.800 me long to connect the dots, and then I can 0:02:00.040 --> 0:02:03.120 rack wherever you go. How did safe graph get this 0:02:03.200 --> 0:02:06.000 information in the first place. Well, it works with app 0:02:06.040 --> 0:02:09.600 developers to get that data. In some cases, safe graph 0:02:09.680 --> 0:02:14.080 relies upon an API or application programming interface. The purpose 0:02:14.120 --> 0:02:16.560 of an API is to allow developers the chance to 0:02:16.600 --> 0:02:19.960 create new apps that can communicate with and take advantage 0:02:20.000 --> 0:02:23.640 of other apps. This allows for a mutually beneficial relationship 0:02:23.680 --> 0:02:28.040 between multiple applications. In other cases, safe graph negotiates a 0:02:28.120 --> 0:02:32.200 price for that information. It's true everywhere, but especially for 0:02:32.280 --> 0:02:37.119 the online world, information is valuable. There are entire companies 0:02:37.160 --> 0:02:40.079 that exist solely to buy and sell massive amounts of 0:02:40.200 --> 0:02:43.360 data about users. And again, while on the surface that 0:02:43.440 --> 0:02:46.280 data may not be associated with a specific name, it 0:02:46.320 --> 0:02:49.040 typically doesn't take a lot of work to associate a 0:02:49.080 --> 0:02:52.279 specific person with a particular device or set of behaviors. 0:02:52.600 --> 0:02:55.880 Studies bear this out. A research paper published in Science 0:02:55.880 --> 0:02:59.639 in showed that by analyzing meta data connected to credit 0:02:59.680 --> 0:03:04.280 card transactions, it was trivial to connect purchases to specific people. 0:03:04.720 --> 0:03:07.520 This was after all personal data had been scrubbed from 0:03:07.560 --> 0:03:11.240 the information. The data sets included dates, the amounts that 0:03:11.280 --> 0:03:13.840 were charged, and the stores that were visited. There were 0:03:13.880 --> 0:03:16.560 no names, no credit card numbers, or any other information 0:03:16.560 --> 0:03:20.120 that overtly identified the card owners. It was the behavior 0:03:20.240 --> 0:03:23.600 of shoppers that made them easy to identify. That, coupled 0:03:23.600 --> 0:03:27.239 with publicly available information such as public posts on social media, 0:03:27.520 --> 0:03:31.120 made it simple to re identify nine of the shoppers 0:03:31.280 --> 0:03:35.200 in the study. This research paper isn't an outlier. There 0:03:35.200 --> 0:03:37.480 have been numerous other studies that have shown that massive 0:03:37.480 --> 0:03:41.640 amounts of supposedly anonymous information give anyone with patients enough 0:03:41.680 --> 0:03:45.920 of a lead to identify specific people among that data set, 0:03:46.200 --> 0:03:49.000 and as you might imagine, this poses a serious threat 0:03:49.040 --> 0:03:52.440 to people's privacy. It doesn't take a worse case scenario 0:03:52.520 --> 0:03:55.280 to see how all this data mining and analysis could 0:03:55.320 --> 0:03:57.840 go wrong. There are plenty of people who would love 0:03:57.880 --> 0:04:00.560 to take advantage of all that information. Some want to 0:04:00.600 --> 0:04:04.840 send targeted advertising to users. In a typical implementation, an 0:04:04.880 --> 0:04:07.720 algorithm would match adds to the people most likely to 0:04:07.760 --> 0:04:11.200 find those ads interesting. Some people might find that approach 0:04:11.280 --> 0:04:13.760 off putting or irritating, but it's one of the more 0:04:13.800 --> 0:04:17.440 benign uses for all that data. Others might take the 0:04:17.480 --> 0:04:20.240 information and use it for all sorts of nefarious purposes. 0:04:20.520 --> 0:04:22.760 It's not a stretch to imagine a situation in which 0:04:22.760 --> 0:04:25.640 someone uses this kind of data to blackmail a target. 0:04:25.920 --> 0:04:28.239 Just imagine getting a phone call and hearing someone say, hey, 0:04:28.520 --> 0:04:31.479 you called out sick to work yesterday, but I see 0:04:31.520 --> 0:04:34.080 you were actually visiting an amusement park all day long. 0:04:34.960 --> 0:04:36.760 Not that I would ever miss work to go to 0:04:36.800 --> 0:04:40.799 an amusement park. Or imagine an insurance company combing through 0:04:40.800 --> 0:04:42.839 this data to figure out how much it should charge 0:04:42.880 --> 0:04:45.640 customers based on their behaviors, or a company using it 0:04:45.680 --> 0:04:47.960 to keep an eye on what employees are doing off 0:04:48.040 --> 0:04:52.200 the clock. It's a massive invasion of privacy. So what's 0:04:52.240 --> 0:04:55.520 the solution? From a personal standpoint? You can turn off 0:04:55.600 --> 0:04:58.520 location features on your devices That will help a bit. 0:04:58.960 --> 0:05:02.280 You can avoid attaching location data to social posts. You 0:05:02.320 --> 0:05:05.760 can turn social posts to private or limit giving information 0:05:05.760 --> 0:05:08.120 about the places you go to and the things you 0:05:08.160 --> 0:05:11.719 do on social media. In short, you can withdraw from 0:05:11.760 --> 0:05:14.320 many of the activities that are becoming more common in 0:05:14.360 --> 0:05:17.880 everyday life. It's not a fun answer, but it is 0:05:17.920 --> 0:05:20.600 an honest one. Even then, there are things that may 0:05:20.640 --> 0:05:25.280 be outside your awareness or control that are being tracked. Ultimately, 0:05:25.360 --> 0:05:28.080 what we might need our regulations on how companies can 0:05:28.120 --> 0:05:31.880 collect and perhaps more importantly, profit from the data we generate. 0:05:32.520 --> 0:05:35.320 Here's hoping you have a happy and safe holiday season. 0:05:35.600 --> 0:05:38.360 If you want to learn more about online privacy, GPS, 0:05:38.400 --> 0:05:41.839 technology and big data, subscribe to The Tech Stuff podcast. 0:05:42.040 --> 0:05:44.680 We explore technology on all scales in a long form 0:05:44.720 --> 0:05:47.719 podcast that publishes every Wednesday and Friday. I'll see you 0:05:47.760 --> 0:05:48.360 again soon.