recording-subtitles/2021/PII/output.srt

1
00:00:00,000 --> 00:00:11,880
Thank you once again for coming for the last talk of the day, at least for you.

2
00:00:11,880 --> 00:00:17,840
With me, Thomas Persson, he's a business developer at Digitalist Sweden, he has worked with tracking

3
00:00:17,840 --> 00:00:25,380
digital analytics since 2010 and has been a contributor to open source since 2007.

4
00:00:25,380 --> 00:00:28,920
What happens under the hood when you visit a website and what data is actually shared

5
00:00:28,920 --> 00:00:31,720
about you and who can actually access this data?

6
00:00:31,720 --> 00:00:34,720
In this talk, Thomas will answer these questions.

7
00:00:34,720 --> 00:00:35,720
Please begin.

8
00:00:35,720 --> 00:00:36,720
All right.

9
00:00:36,720 --> 00:00:37,720
Thank you.

10
00:00:37,720 --> 00:00:42,520
Yeah, I'm getting a bit tired now.

11
00:00:42,520 --> 00:00:50,680
Last session, for me at least, for this day, I'm really looking forward to having a long

12
00:00:50,680 --> 00:00:56,600
weekend after this, but let's push it over the finish line.

13
00:00:56,600 --> 00:01:04,760
So in this presentation, I will talk a little bit more on a higher level about privacy on

14
00:01:04,760 --> 00:01:11,040
the web and what type of data and try to explain what happens under the hood when we're visiting

15
00:01:11,040 --> 00:01:20,320
websites and also talk about personal identifying information in relation to web analytics and

16
00:01:20,320 --> 00:01:27,000
then finally give you some examples of how to manage this in Atom.

17
00:01:27,000 --> 00:01:29,240
So let's get started.

18
00:01:29,240 --> 00:01:37,320
Let's see if I can just find that correct window.

19
00:01:37,320 --> 00:01:41,400
So yeah, I got the presentation already, so I will skip that.

20
00:01:41,400 --> 00:01:48,720
You probably know Digitalist at the time, so I will skip that so we can move forward.

21
00:01:48,720 --> 00:01:55,520
But here's the kind of issues we have right now with privacy and the reason why I think

22
00:01:55,520 --> 00:02:01,560
a lot of organizations are moving away from tools like Google Analytics, but basically

23
00:02:01,560 --> 00:02:12,320
any cloud provider that is not provided by the EU and trying to do the simple explanation

24
00:02:12,320 --> 00:02:13,440
here at least.

25
00:02:13,440 --> 00:02:15,160
So this is me.

26
00:02:15,160 --> 00:02:16,920
I'm a citizen in Sweden.

27
00:02:16,920 --> 00:02:19,080
I live in Stockholm.

28
00:02:19,080 --> 00:02:28,400
So if I go to stockholm.se where I can find a lot of information to me as a citizen.

29
00:02:28,400 --> 00:02:35,440
I live in Sweden, so that means that we have a law in Sweden called OSL.

30
00:02:35,440 --> 00:02:41,400
It's a Swedish law about publicity and secrecy.

31
00:02:41,400 --> 00:02:46,040
It's been in place for quite some time and that law applies to me as a Swedish citizen.

32
00:02:46,040 --> 00:02:50,480
And it applies to the organization running stockholm.se.

33
00:02:50,480 --> 00:02:56,200
We also have GDPR, which is a European law and that applies to both of us as well since

34
00:02:56,200 --> 00:03:04,840
EU is the level above us, so to speak.

35
00:03:04,840 --> 00:03:10,800
To provide this service, stockholm.se probably has a lot of service providers.

36
00:03:10,800 --> 00:03:15,840
They probably have a company doing their hosting, their development and so on.

37
00:03:15,840 --> 00:03:17,880
Maybe design work or whatever.

38
00:03:17,880 --> 00:03:24,520
So a lot of actors are involved in this work and they all apply to these laws.

39
00:03:24,520 --> 00:03:30,680
So when stockholm.se write an agreement with these companies, these laws apply.

40
00:03:30,680 --> 00:03:34,200
OSL and GDPR applies to these contracts.

41
00:03:34,200 --> 00:03:38,960
The problem though is that when you start to write contracts with companies that are

42
00:03:38,960 --> 00:03:45,800
not part of the European Union, like for instance if you use Office 365 or if you sign up for

43
00:03:45,800 --> 00:03:54,040
Google Analytics on your site, the problem here is that Microsoft or Google or other

44
00:03:54,040 --> 00:04:01,760
companies or basically any cloud provider, they live under other laws.

45
00:04:01,760 --> 00:04:09,680
And in this particular example, we have something called the FISA 702 and that means that even

46
00:04:09,680 --> 00:04:15,920
though we write a contract with Microsoft, these laws will override that and Microsoft

47
00:04:15,920 --> 00:04:19,400
can't do anything about that, even though they would like to.

48
00:04:19,400 --> 00:04:26,960
So this is kind of the main problem with using cloud services and what FISA 702 actually

49
00:04:26,960 --> 00:04:39,640
says here is that if any American governmental institution basically requests for data, they

50
00:04:39,640 --> 00:04:47,760
in a Microsoft facility, they have to give it to that organization and they can't stop

51
00:04:47,760 --> 00:04:51,640
that in any way, at least not right now.

52
00:04:51,640 --> 00:04:57,400
So this means that any data that is stored in a facility owned by Microsoft or Google

53
00:04:57,400 --> 00:05:06,120
or whatever, and it doesn't matter if their physical servers are in Europe, you have to

54
00:05:06,120 --> 00:05:13,180
think about it as if it's readable by the American public organizations and we'll look

55
00:05:13,180 --> 00:05:17,880
into a little bit more about the problems with this later on.

56
00:05:17,880 --> 00:05:22,360
So we will also look into the actual details.

57
00:05:22,360 --> 00:05:27,320
So Google Analytics is obviously what we discuss when we talk about Matomo.

58
00:05:27,320 --> 00:05:32,880
So yes, we need to think about when you use Google Analytics that we are leaking data

59
00:05:32,880 --> 00:05:41,120
to foreign powers like NSA and so on.

60
00:05:41,120 --> 00:05:46,320
We also have to think about that we're actually leaking personal data without having a consent

61
00:05:46,320 --> 00:05:50,080
and this is an issue with GDPR in Europe.

62
00:05:50,080 --> 00:05:55,840
We're also selling data about our visitors and that's more related to Google specifically.

63
00:05:55,840 --> 00:06:03,800
But historically, this started with quite big, when the whistleblower Edward Snowden

64
00:06:03,800 --> 00:06:10,040
with that didn't start though then, but it actually got really, really clear to us that

65
00:06:10,040 --> 00:06:19,960
when he shared the NSA documents about what was happening, people started to realize how

66
00:06:19,960 --> 00:06:22,200
big this really was.

67
00:06:22,200 --> 00:06:29,040
So we talked about the law called FISA, it's actually called Foreign Intelligence Violence

68
00:06:29,040 --> 00:06:30,040
Act.

69
00:06:30,040 --> 00:06:38,320
It was a law created in the USA to actually protect the American citizen because there

70
00:06:38,320 --> 00:06:44,760
has been a lot of surveillance going on, phone surveillance mainly and other things without

71
00:06:44,760 --> 00:06:47,360
any regulations in the United States.

72
00:06:47,360 --> 00:06:50,960
It was actually created to protect American citizens.

73
00:06:50,960 --> 00:07:00,640
9-11 was the event that happened and that really changed how the American states started

74
00:07:00,640 --> 00:07:02,200
to act differently.

75
00:07:02,200 --> 00:07:08,320
So people at NSA has described this in many, many places that there is a time before and

76
00:07:08,320 --> 00:07:13,560
after 9-11 and we'll see a little bit more about the statistics.

77
00:07:13,560 --> 00:07:22,240
They basically changed the laws to make it even easier to supervise the society and in

78
00:07:22,240 --> 00:07:26,680
particular people that are not members of the United States.

79
00:07:26,680 --> 00:07:30,160
So Europe is a target market for this.

80
00:07:30,160 --> 00:07:33,560
There are a lot of systems in place at NSA.

81
00:07:33,560 --> 00:07:39,960
Some examples are systems called Upstream or Prism and Treasure Map.

82
00:07:39,960 --> 00:07:48,680
This is basically systems that are snooping on internet traffic, phones, and etc.

83
00:07:48,680 --> 00:07:53,920
And they have direct access to data centers at Google, Microsoft, Amazon, etc.

84
00:07:53,920 --> 00:07:58,400
So it's really a large scale.

85
00:07:58,400 --> 00:08:06,400
Also we have to know that internet traffic is passing through United American borders

86
00:08:06,400 --> 00:08:11,760
and this means that it often gets stuck in this system called Upstream.

87
00:08:11,760 --> 00:08:17,720
We'll look into more details about how easily this actually happens and what type of data

88
00:08:17,720 --> 00:08:20,320
they can actually tap into.

89
00:08:20,320 --> 00:08:25,680
So is this really a problem in a country like Sweden?

90
00:08:25,680 --> 00:08:33,360
Is this really something that government agencies use to approach Swedish citizens?

91
00:08:33,360 --> 00:08:36,000
So Sweden is a rather small country in Europe.

92
00:08:36,000 --> 00:08:39,680
We have about 10 million citizens.

93
00:08:39,680 --> 00:08:46,000
But the interesting part here is that Google, Amazon, Microsoft, and a lot of other companies,

94
00:08:46,000 --> 00:08:51,000
I think it was about 50 companies in the beginning after the Snowden incident, have started to

95
00:08:51,000 --> 00:09:01,880
report how often they leave data to public agencies in the States.

96
00:09:01,880 --> 00:09:04,240
So this is the data from Google.

97
00:09:04,240 --> 00:09:08,820
And as you can see, the growth here is pretty massive.

98
00:09:08,820 --> 00:09:21,920
So they have left out information about 235,000 accounts the first six months of 2020.

99
00:09:21,920 --> 00:09:27,600
The data is a bit lacking behind, but the numbers are actually growing.

100
00:09:27,600 --> 00:09:33,960
So the same period in Sweden, we had 370 cases from Google where they handed out information

101
00:09:33,960 --> 00:09:40,200
and you can think about this as, okay, maybe this is information that the Swedish police

102
00:09:40,200 --> 00:09:43,160
is asking from NSA to give them.

103
00:09:43,160 --> 00:09:46,400
But that data is actually not included here.

104
00:09:46,400 --> 00:09:54,140
This is only data requested that American organizations ask for.

105
00:09:54,140 --> 00:09:59,480
So we don't know what this is and it's totally out of our control.

106
00:09:59,480 --> 00:10:07,760
I did a research and checked for Google, Microsoft, and Apple and as you can see, there's quite

107
00:10:07,760 --> 00:10:10,120
a lot of information handed out.

108
00:10:10,120 --> 00:10:16,600
And if you summarize this, it's actually eight times a day data is requested about Swedish

109
00:10:16,600 --> 00:10:19,320
citizens from these four companies.

110
00:10:19,320 --> 00:10:24,720
So there's no doubt that this is a huge, huge problem.

111
00:10:24,720 --> 00:10:31,400
And because of this, we need to change the way, and I think this is extra critical if

112
00:10:31,400 --> 00:10:40,520
you're a public organization handling citizen data or whatever, it goes for private companies

113
00:10:40,520 --> 00:10:43,320
as well, but super important.

114
00:10:43,320 --> 00:10:52,520
So what I usually say is that the public sector needs to secure our infrastructure.

115
00:10:52,520 --> 00:10:59,280
We need to use open technologies such as Matomo, but that's just one thing of the problem.

116
00:10:59,280 --> 00:11:04,000
We need to secure the traffic between the residents and the authorities so it never

117
00:11:04,000 --> 00:11:05,000
leaves Europe.

118
00:11:05,000 --> 00:11:10,720
This is another problem that we saw that if the internet traffic just passes through United

119
00:11:10,720 --> 00:11:25,240
States, it's much probable that it gets stuck in this surveillance network.

120
00:11:25,240 --> 00:11:29,880
So let's go into details of the web.

121
00:11:29,880 --> 00:11:34,560
So what happens when I visit five websites under the hood?

122
00:11:34,560 --> 00:11:40,640
So I did a little animation with a plugin called Lightbeam for Firefox.

123
00:11:40,640 --> 00:11:46,400
So what we're seeing here is what happens, these are requests that's happened under the

124
00:11:46,400 --> 00:11:54,240
hood, and this is actually public sector websites are related in Sweden, and you can see the

125
00:11:54,240 --> 00:12:01,480
spider diagram on the left that all of them under the hood request information from Google

126
00:12:01,480 --> 00:12:02,480
in this case.

127
00:12:02,480 --> 00:12:04,880
They're using Google Analytics basically.

128
00:12:04,880 --> 00:12:12,720
So that means that Google can profile all these visits not only on these particular

129
00:12:12,720 --> 00:12:23,160
websites, but they can create profiles on individual level of pretty sensitive sites.

130
00:12:23,160 --> 00:12:27,880
So yes, they use Google Analytics, and we need to remember that this is something that

131
00:12:27,880 --> 00:12:29,440
Google sells, of course.

132
00:12:29,440 --> 00:12:37,040
The 3% of the revenue two years ago, I think, was coming from ad sales basically.

133
00:12:37,040 --> 00:12:41,700
So it's a huge problem.

134
00:12:41,700 --> 00:12:47,640
So it's not just about the privacy and that they're leaking data, we're actually selling

135
00:12:47,640 --> 00:12:50,920
their behavior to Google.

136
00:12:50,920 --> 00:12:54,800
And Google is just one part of the problem.

137
00:12:54,800 --> 00:12:58,800
You need to be aware of everything you do on your website.

138
00:12:58,800 --> 00:13:04,520
So this is another Swedish, it's actually the government website in Sweden, and I've

139
00:13:04,520 --> 00:13:08,240
been picking on them for a few years now.

140
00:13:08,240 --> 00:13:11,240
The good thing is that they have actually changed this now.

141
00:13:11,240 --> 00:13:16,520
So if you go to this site, they actually change how it works, and I think I affected them

142
00:13:16,520 --> 00:13:17,920
a bit.

143
00:13:17,920 --> 00:13:20,160
But the example is still good.

144
00:13:20,160 --> 00:13:26,800
So if you go to a page called contact on the Swedish government site, and you look under

145
00:13:26,800 --> 00:13:33,240
the hood, you do a request in your browser, and it's sent to their server, and back we

146
00:13:33,240 --> 00:13:34,240
get HTML.

147
00:13:34,240 --> 00:13:35,240
You all know this.

148
00:13:35,240 --> 00:13:44,040
And in this HTML, we're actually requesting 33 other resources from six different domains.

149
00:13:44,040 --> 00:13:50,160
And for each of these requests, the way internet works is actually we're sending my IP address

150
00:13:50,160 --> 00:13:52,680
from my browser to every server.

151
00:13:52,680 --> 00:14:00,560
And if we dig down in this and look into a specific one of these requests, it's a little

152
00:14:00,560 --> 00:14:09,240
JavaScript called find.js, and it's coming from this domain, dl.episerver.net.

153
00:14:09,240 --> 00:14:15,280
And what's interesting here is that with this request, we're actually also sending something

154
00:14:15,280 --> 00:14:17,000
called the header.

155
00:14:17,000 --> 00:14:21,880
The header contains another information that is called the referrer.

156
00:14:21,880 --> 00:14:27,720
We're also having information such as the language in my browser.

157
00:14:27,720 --> 00:14:33,360
We can see what browser I have in the user agent, et cetera, et cetera.

158
00:14:33,360 --> 00:14:39,880
So basically the information shared with this server, dl.episerver.net, is the same kind

159
00:14:39,880 --> 00:14:46,120
of information that you use when you track data in Matomo, meaning all the sites that

160
00:14:46,120 --> 00:14:51,760
we request for information just to present our website is basically able to do the same

161
00:14:51,760 --> 00:14:56,780
profiling that we can do with Matomo or any other tool on our website.

162
00:14:56,780 --> 00:15:03,120
So you really need to control what type of resources you include on your websites, and

163
00:15:03,120 --> 00:15:08,120
they have to be within the European Union if you're a European company.

164
00:15:08,120 --> 00:15:09,440
All right.

165
00:15:09,440 --> 00:15:16,200
So there are some web standards where you can actually try to block this.

166
00:15:16,200 --> 00:15:23,960
There's something called the referrer policy, and if you set this to no referrer, you actually

167
00:15:23,960 --> 00:15:27,480
can block sending the referrer information.

168
00:15:27,480 --> 00:15:34,080
The problem though is that this standard is not respected by Microsoft, so all Microsoft

169
00:15:34,080 --> 00:15:40,720
browsers will send that anyway, and it's also not respected by iPhones.

170
00:15:40,720 --> 00:15:45,640
So Safari and iPhones will send that even though you try to block it, meaning quite

171
00:15:45,640 --> 00:15:55,080
a few percentages of the visitors you will have on your website since it's quite common

172
00:15:55,080 --> 00:15:57,760
software combinations or hardware combinations.

173
00:15:57,760 --> 00:15:58,760
All right.

174
00:15:58,760 --> 00:16:02,880
So to summarize, we do 26 calls.

175
00:16:02,880 --> 00:16:09,960
Eight of these calls goes to four different domains in the United States, and in all these

176
00:16:09,960 --> 00:16:15,480
we share the IP address and the URL, meaning that all these domains can basically start

177
00:16:15,480 --> 00:16:18,480
profiling our visitors.

178
00:16:18,480 --> 00:16:21,800
All right.

179
00:16:21,800 --> 00:16:29,180
So the next thing, because that's really about the URLs and the IP addresses that are sensitive

180
00:16:29,180 --> 00:16:34,940
information that we need to manage and handle in our Matomo installation, for instance.

181
00:16:34,940 --> 00:16:39,400
But what about personal data in general?

182
00:16:39,400 --> 00:16:45,680
So personal data can be defined as simple things as IP addresses, as we mentioned before,

183
00:16:45,680 --> 00:16:51,680
email addresses, zip codes, social security numbers, et cetera, et cetera.

184
00:16:51,680 --> 00:16:53,900
This list you've probably seen.

185
00:16:53,900 --> 00:16:59,640
This means that when we're setting up tracking, we do not want to store these things in Matomo.

186
00:16:59,640 --> 00:17:07,800
And usually this is not something people do, even though there are issues, of course, sometimes.

187
00:17:07,800 --> 00:17:17,320
But more problematic is the thing, the fact that data in combination with other data that

188
00:17:17,320 --> 00:17:24,120
identifies an individual is also to be considered as personal data.

189
00:17:24,120 --> 00:17:31,360
And the best example I have on this is when you start to store location information.

190
00:17:31,360 --> 00:17:38,520
So in this example, I have a little village in Sweden with 55th inhabitants.

191
00:17:38,520 --> 00:17:39,520
It's called Lomtresk.

192
00:17:39,520 --> 00:17:43,160
So now you know a bit of Swedish as well.

193
00:17:43,160 --> 00:17:48,400
The interesting part here is if you store that dimension of data together with your

194
00:17:48,400 --> 00:17:53,200
page views, you can start to get problems pretty quickly.

195
00:17:53,200 --> 00:18:00,480
So think about the other meta information or metadata you store in analytics or in your

196
00:18:00,480 --> 00:18:03,600
Matomo installation.

197
00:18:03,600 --> 00:18:10,680
The first one would be, like, let's say we have a custom dimension with the profession.

198
00:18:10,680 --> 00:18:13,980
Or maybe someone has been searching for a job.

199
00:18:13,980 --> 00:18:19,400
Or something that we can identify a job role here.

200
00:18:19,400 --> 00:18:26,880
Or we can identify that this is the language is Turkish in the browser.

201
00:18:26,880 --> 00:18:29,320
And maybe we can identify that this is a woman.

202
00:18:29,320 --> 00:18:35,120
We can think about how quickly, how easy it starts to get to identify this person just

203
00:18:35,120 --> 00:18:41,440
from these quite anonymous combinations of data.

204
00:18:41,440 --> 00:18:45,160
We can continue.

205
00:18:45,160 --> 00:18:51,360
Just an unusual mobile phone would actually be a problematic thing here.

206
00:18:51,360 --> 00:18:56,760
Maybe everyone in Lomtresk knows who has an iPhone of a specific model.

207
00:18:56,760 --> 00:19:04,640
Or you can find that in an image on Facebook or whatever and start to identify visitors

208
00:19:04,640 --> 00:19:05,640
here.

209
00:19:05,640 --> 00:19:11,320
So that leads us to that personal data is not something fixed.

210
00:19:11,320 --> 00:19:19,600
We need to continuously monitor this and we need to continuously be able to act.

211
00:19:19,600 --> 00:19:26,680
And what we also need to know is that when we include other services on our website,

212
00:19:26,680 --> 00:19:29,760
visitors sharing this information with them.

213
00:19:29,760 --> 00:19:36,840
So if you add, most of them are big in Sweden, but it's a lot of things that you need to

214
00:19:36,840 --> 00:19:41,600
look into and make sure that you're not sharing data with.

215
00:19:41,600 --> 00:19:44,320
I actually created a little tool.

216
00:19:44,320 --> 00:19:45,520
It's actually a JavaScript.

217
00:19:45,520 --> 00:19:52,720
So if you go to my GitHub profile, what this one does is that I'm actually going to show

218
00:19:52,720 --> 00:19:54,480
you.

219
00:19:54,480 --> 00:20:03,440
So if I copy this code and I can basically do this on any website.

220
00:20:03,440 --> 00:20:12,120
So let's do it on atomocamp.org and see if we are okay here.

221
00:20:12,120 --> 00:20:14,200
I'm going to open the inspect.

222
00:20:14,200 --> 00:20:22,600
We're going to just paste this little script and what it will do is to start a request.

223
00:20:22,600 --> 00:20:26,360
Okay, this one was pretty good.

224
00:20:26,360 --> 00:20:33,600
We only included one external script and the country it was requesting was Germany.

225
00:20:33,600 --> 00:20:43,480
So let's do this at the Swedish government again and see how well they are behaving.

226
00:20:43,480 --> 00:20:49,880
I'm going to go there and I'm going to paste the JavaScript and now we're requesting a

227
00:20:49,880 --> 00:20:57,040
lot of websites and I can actually see that they improved since I picked on them.

228
00:20:57,040 --> 00:21:03,240
From all of these sites, we are pulling different scripts and here you can see the country from

229
00:21:03,240 --> 00:21:04,960
where they are requested.

230
00:21:04,960 --> 00:21:08,880
So they did pretty good.

231
00:21:08,880 --> 00:21:14,720
That was not the scenario six months ago and this is good because this means that even

232
00:21:14,720 --> 00:21:23,280
though they share data with these players, they're even in Sweden here so if they have

233
00:21:23,280 --> 00:21:30,360
a contract with them not to share data with anyone else, we're actually legally safe.

234
00:21:30,360 --> 00:21:38,440
Which wouldn't be the case if the country of the scripts would be the United States

235
00:21:38,440 --> 00:21:44,560
and one very common thing is to use CDN networks to load JavaScript for instance and that would

236
00:21:44,560 --> 00:21:50,120
be a problem.

237
00:21:50,120 --> 00:21:54,520
So what do we do when the accident occurs?

238
00:21:54,520 --> 00:22:00,960
So since storing personal data is something you should actually plan for that you will

239
00:22:00,960 --> 00:22:07,360
do, that's usually how I talk to my clients.

240
00:22:07,360 --> 00:22:11,320
You can't really stop this from happening.

241
00:22:11,320 --> 00:22:15,400
You need to plan for when it happens.

242
00:22:15,400 --> 00:22:24,080
So you have to have routines in place to act when you find personal data in your analytics

243
00:22:24,080 --> 00:22:28,120
and this is the action plan I usually show.

244
00:22:28,120 --> 00:22:32,200
So first we need to assess how severe this is.

245
00:22:32,200 --> 00:22:38,260
I mean how sensitive is the data that we have leaked?

246
00:22:38,260 --> 00:22:42,480
As quickly as possible we prevent further collection.

247
00:22:42,480 --> 00:22:46,480
So data is not collected more than necessary.

248
00:22:46,480 --> 00:22:55,000
In Sweden we actually have to decide within 72 hours after the incident, we notice the

249
00:22:55,000 --> 00:23:05,600
incident to our reporting organization basically for GDPR and data leakage.

250
00:23:05,600 --> 00:23:11,080
What we need to do after this is of course to try to clear or anonymize the data.

251
00:23:11,080 --> 00:23:21,920
We need to document the incident and potentially inform the victims like if users' data was

252
00:23:21,920 --> 00:23:27,920
really sensitive we might have to contact them even and then I usually say that let's

253
00:23:27,920 --> 00:23:32,760
have a retrospective to see if we can avoid that this happens again.

254
00:23:32,760 --> 00:23:40,880
Maybe this is about training because most of these mistakes are personal but not always.

255
00:23:40,880 --> 00:23:44,000
Sometimes there are technical solutions.

256
00:23:44,000 --> 00:23:52,500
Clearing data though is something that we need to work pretty much with and that's actually

257
00:23:52,500 --> 00:23:59,400
something that is possible if you're using Matomo but it's not necessarily easy just

258
00:23:59,400 --> 00:24:01,320
because you can.

259
00:24:01,320 --> 00:24:04,520
So deleting data in Matomo usually looks like this.

260
00:24:04,520 --> 00:24:09,680
First of all you have to find the data, you have to identify the actual visitors and the

261
00:24:09,680 --> 00:24:10,680
raw data.

262
00:24:10,680 --> 00:24:14,160
So you have to find the visitor IDs that are affected.

263
00:24:14,160 --> 00:24:21,120
You need to delete these visits, I will show you some examples soon and then you need to

264
00:24:21,120 --> 00:24:26,080
reprocess your visitor log so that the aggregated data reports are updated.

265
00:24:26,080 --> 00:24:34,520
So for instance if you find an email address in your page view report you would first need

266
00:24:34,520 --> 00:24:43,520
to find the visitors that sent this data in and delete those and then you need to reprocess

267
00:24:43,520 --> 00:24:50,160
your aggregated reports for the dates when that data was stored.

268
00:24:50,160 --> 00:24:58,240
So that's quite a big job to do that.

269
00:24:58,240 --> 00:25:04,040
One thing you can do to start monitoring this is actually to use a plugin called the alert

270
00:25:04,040 --> 00:25:13,680
plugin and in that plugin you can start to set up rules to identify personal sensitive

271
00:25:13,680 --> 00:25:18,200
information and I will show you an example of this.

272
00:25:18,200 --> 00:25:23,680
So what you do is that you for instance if you want to start monitoring your page views

273
00:25:23,680 --> 00:25:27,920
you can set up an alert and you can use the regular expression.

274
00:25:27,920 --> 00:25:36,040
In the example here I'm matching a date but I will give you some and then if we have more

275
00:25:36,040 --> 00:25:43,840
than one page view containing a date I will get an alarm or a report that at least highlights

276
00:25:43,840 --> 00:25:45,920
that I need to look at this.

277
00:25:45,920 --> 00:25:50,880
So an example here is that you can find a lot of regular expressions online.

278
00:25:50,880 --> 00:25:54,080
So in this case I have the example of an email address.

279
00:25:54,080 --> 00:26:00,000
So let's monitor if we have email addresses in our events or our page views and ping me

280
00:26:00,000 --> 00:26:04,800
if that happens.

281
00:26:04,800 --> 00:26:10,080
Finally you can combine these into a long string so you don't have to create individual

282
00:26:10,080 --> 00:26:14,440
reports for every monitor you have.

283
00:26:14,440 --> 00:26:20,400
The drawback of this is that the alert plugin will only process once a day or every night.

284
00:26:20,400 --> 00:26:28,560
You can't do that more often and it's also hard to identify something like a name or

285
00:26:28,560 --> 00:26:29,560
something like that.

286
00:26:29,560 --> 00:26:38,400
So it's only data that is quite easy to identify that you can find this way.

287
00:26:38,400 --> 00:26:45,960
Once you find it you can use a tool called GDPR tools and try to delete things that way.

288
00:26:45,960 --> 00:26:53,440
You need to find the visitor ID and when you do so you can delete that.

289
00:26:53,440 --> 00:26:57,760
You can also use the API, the example below.

290
00:26:57,760 --> 00:27:11,120
If you have a lot of data you can create a script and call the API many times.

291
00:27:11,120 --> 00:27:14,440
I can also show you what is a bit problematic.

292
00:27:14,440 --> 00:27:22,240
So in this case I look at the pages report and now I'm just simulating but let's say

293
00:27:22,240 --> 00:27:27,460
that this is my name on our site so it's not really personal data.

294
00:27:27,460 --> 00:27:33,520
But what if I wanted to delete this information because it's sensitive?

295
00:27:33,520 --> 00:27:43,800
So what I would do is to click on this one, segmented visitor log.

296
00:27:43,800 --> 00:27:51,200
So when I click this one I will actually get the visits and in this case I even have a

297
00:27:51,200 --> 00:27:57,240
little JavaScript in my browser so I actually have a little button here that allows me to

298
00:27:57,240 --> 00:27:59,320
delete this directly.

299
00:27:59,320 --> 00:28:05,040
But otherwise you can hover on the IP address and you can actually find the visitor ID up

300
00:28:05,040 --> 00:28:06,040
there.

301
00:28:06,040 --> 00:28:07,600
You can probably see it in the screen.

302
00:28:07,600 --> 00:28:15,360
This is the information that you need to grab and then use in the GDPR tools and apply

303
00:28:15,360 --> 00:28:16,360
it here.

304
00:28:16,360 --> 00:28:23,000
It's called visitor ID.

305
00:28:23,000 --> 00:28:25,520
You can actually steal this little number.

306
00:28:25,520 --> 00:28:29,280
So as you see it's starting to get quite time consuming.

307
00:28:29,280 --> 00:28:35,560
If you have a lot of personal data this is a terribly time consuming job.

308
00:28:35,560 --> 00:28:39,560
This little JavaScript helps me.

309
00:28:39,560 --> 00:28:44,440
What it actually does is that it calls the API.

310
00:28:44,440 --> 00:28:47,440
I have just created a link to that.

311
00:28:47,440 --> 00:29:01,380
It's actually this little line here and I automatically add the ID visit to delete it.

312
00:29:01,380 --> 00:29:10,640
So the idea I have for the future is to actually build a PII monitoring plugin for Matomo.

313
00:29:10,640 --> 00:29:18,560
That plugin would first of all set up some predefined rules to identify things like email

314
00:29:18,560 --> 00:29:24,800
addresses, credit card numbers, et cetera, so that we have alarms.

315
00:29:24,800 --> 00:29:31,280
I would also set it up to watch more types of data directly like events, page views,

316
00:29:31,280 --> 00:29:32,280
refer URLs.

317
00:29:32,280 --> 00:29:36,720
You can come up with a lot of ideas.

318
00:29:36,720 --> 00:29:40,600
I would also make it possible to add exceptions.

319
00:29:40,600 --> 00:29:49,320
After you try to identify something or identify something that is actually okay.

320
00:29:49,320 --> 00:29:52,280
So we want to add an exception to that.

321
00:29:52,280 --> 00:29:56,960
I would also like to make it easy to delete data so the report would actually give me

322
00:29:56,960 --> 00:30:04,520
all the visitor IDs and make it possible to delete them directly.

323
00:30:04,520 --> 00:30:10,240
I would also like to be possible to just anonymize the data and not maybe delete it.

324
00:30:10,240 --> 00:30:15,760
So let's say I found a credit card number but I still want to have the traffic stored

325
00:30:15,760 --> 00:30:16,760
in my website.

326
00:30:16,760 --> 00:30:25,880
I just want to anonymize that information.

327
00:30:25,880 --> 00:30:31,200
But that's not really possible today unless you write SQL scripts.

328
00:30:31,200 --> 00:30:40,720
So these are kind of the ideas that I have around this plugin.

329
00:30:40,720 --> 00:30:43,160
It shouldn't be too hard to do the basics.

330
00:30:43,160 --> 00:30:48,040
The hard part is of course to start monitoring combinations of data.

331
00:30:48,040 --> 00:30:57,560
Like it would be impossible almost to do what I showed you with the small village Lundträsk

332
00:30:57,560 --> 00:31:01,680
and identify those individuals.

333
00:31:01,680 --> 00:31:07,560
Alright, so the last slide actually or the second last slide.

334
00:31:07,560 --> 00:31:12,480
But usually this is what I recommend people to look into if you want to secure your web

335
00:31:12,480 --> 00:31:14,000
analysis.

336
00:31:14,000 --> 00:31:19,160
So first of all collect consents from your visitor and inform them that you are tracking.

337
00:31:19,160 --> 00:31:22,080
This is all from the GDPR.

338
00:31:22,080 --> 00:31:23,840
Check your data collection.

339
00:31:23,840 --> 00:31:29,440
Find the obvious collection of personal identification information and set up your Matomo instance

340
00:31:29,440 --> 00:31:30,800
properly.

341
00:31:30,800 --> 00:31:39,200
That means anonymizing IP addresses, not storing data longer than needed, etc, etc.

342
00:31:39,200 --> 00:31:43,280
Also make sure to manage the quality of your application.

343
00:31:43,280 --> 00:31:49,920
So they are not sending personal identification information because that's normally one of

344
00:31:49,920 --> 00:31:52,800
the issues we have when we get bad data.

345
00:31:52,800 --> 00:31:56,280
It's often the applications that are wrong.

346
00:31:56,280 --> 00:32:00,000
Also make sure to secure your technical infrastructure.

347
00:32:00,000 --> 00:32:05,680
The service has to be within Europe and it has to be owned by a European company.

348
00:32:05,680 --> 00:32:12,080
You can't use cloud providers in Europe because you're invalidating GDPR by doing so.

349
00:32:12,080 --> 00:32:17,200
Also make sure to limit the time for how long you store data.

350
00:32:17,200 --> 00:32:25,440
Make sure to have processes in place to monitor data and make sure that you know how to act

351
00:32:25,440 --> 00:32:28,840
when you find problems.

352
00:32:28,840 --> 00:32:31,920
Alright, so that was the last slide.

353
00:32:31,920 --> 00:32:37,560
I think it's time for questions.

354
00:32:37,560 --> 00:32:40,040
Thank you for this very interesting talk.

355
00:32:40,040 --> 00:32:48,160
If there are any questions, please post them or write them down in the corresponding chat.

356
00:32:48,160 --> 00:32:55,440
There's been some small discussion of the privacy test and I will share it if it's alright

357
00:32:55,440 --> 00:32:56,440
for you.

358
00:32:56,440 --> 00:32:57,440
Yeah, I saw that.

359
00:32:57,440 --> 00:32:58,440
Nice, Lukas.

360
00:32:58,440 --> 00:33:05,240
That one, of course, is a good thing.

361
00:33:05,240 --> 00:33:09,520
So who wants to help me write the PII plug-in, Lukas?

362
00:33:09,520 --> 00:33:17,960
Are you ready to help?

363
00:33:17,960 --> 00:33:18,960
Check my calendar.

364
00:33:18,960 --> 00:33:19,960
Yeah, of course.

365
00:33:19,960 --> 00:33:22,960
It's the same here, I think.

366
00:33:22,960 --> 00:33:28,640
But it's definitely just something I think a lot of people would appreciate.

367
00:33:28,640 --> 00:33:35,720
Alright, so if we don't have more questions, I'm going to say thank you for today and hope

368
00:33:35,720 --> 00:33:36,720
you enjoyed it.

369
00:33:36,720 --> 00:33:41,400
I'm pretty tired now, so I'm going to have a really nice weekend with my family, maybe

370
00:33:41,400 --> 00:33:48,880
go out and pick some mushrooms and yeah, be outside.

371
00:33:48,880 --> 00:33:53,840
There would be one last question that you would be willing to answer.

372
00:33:53,840 --> 00:33:59,280
Is there any other kind of information apart from cookies and JavaScript code, codes sent

373
00:33:59,280 --> 00:34:02,480
to third parties?

374
00:34:02,480 --> 00:34:10,800
Anything you basically include, so it would be fonts, images, anything.

375
00:34:10,800 --> 00:34:15,600
Anything you request might actually share that information.

376
00:34:15,600 --> 00:34:21,080
Thank you for that.

377
00:34:21,080 --> 00:34:28,080
And if there aren't any other questions, which I think there aren't, then yes, enjoy a long

378
00:34:28,080 --> 00:34:29,080
weekend.

379
00:34:29,080 --> 00:34:30,080
Yes, thank you.

380
00:34:30,080 --> 00:34:31,080
Let's see.

381
00:34:31,080 --> 00:34:38,080
Maybe more questions coming, we'll see.

382
00:34:38,080 --> 00:34:42,760
Oh yeah, a ton of them.

383
00:34:42,760 --> 00:34:46,240
Maybe they're just thanking you for the great talks.

384
00:34:46,240 --> 00:34:49,240
We'll see.

385
00:34:49,240 --> 00:34:54,360
Two seconds.

386
00:34:54,360 --> 00:34:58,160
It seems to be the case that there aren't any more questions coming.

387
00:34:58,160 --> 00:34:59,160
Let's have a nice weekend.

388
00:34:59,160 --> 00:35:00,160
That sounds good.

389
00:35:00,160 --> 00:35:01,160
So let's call it here.

390
00:35:01,160 --> 00:35:02,160
Yeah.

391
00:35:02,160 --> 00:35:07,480
As always, if there are any other questions, you can still be asked.

392
00:35:07,480 --> 00:35:10,600
And yeah, have a great day, have a great weekend.

393
00:35:10,600 --> 00:35:30,920
Thanks for your talks.