 Talk, fake news, how to fight it and how to make sure that you get your information right. By Tony and King BBQ. We'd be very grateful for feedback on Twitter using the hashtag hashtag C3T or at C3 lingo. Thank you. So Harold is just introducing the speaker. So who of you has used image research before to find out whether you've seen a picture before or... Okay, very good. That's about a third. Now, I actually did this quite recently and it's quite useful. And especially right now where oftentimes you hear fake news or live press. It's really, really important to find out what's real and what isn't real. And we know that you can actually have deep fakes so it's becoming easier and easier to produce fake news. And today we have an expert here who does production for a TV show and how verification teams work there. Thank you and applause for Robert. Yeah, thank you. Nice that you guys made it even though it's quite late. And I was quite glad to see how many of you have already done reverse image search. That kind of makes me hope. And now I want to explain what further possibilities and tools there are to make it easier to verify sort of images. And actually verification isn't rocket science. Everyone can do it. And now I'll explain why I don't really like fake news but it can be quite useful or it can be the right term. And then I'll talk about what kind of fake stuff we have, why people are faking stuff. And then I have a couple of examples to show how you can actually find out that they falsified it. Now let's start defining fake news. Our wrong information is wrong information that is often sent out via social media, via electronic channels. They often start from an individual or a group that I have been asked to do that. Now of course this many different fake news is basically a huge term that includes many, many different terms. Now before we had fake news, we basically had different terms for all of these things. And a lie is still a lie. It's not fake news actually. And this fake news does not have any information about what kind of falsifying it is, who made it and the motivation. And those are the kind of information that are actually important for us. Now making fakes and sending them out, publishing them, has never been as easy as it is today. Of course, we've had internet access and smart phones for a while now. But other parts of the world have only recently gotten smart phones and also in crisis regions. Especially when we look at Syria, that's sort of the first war that we have as much pictures, as many pictures, and so many pictures that we can't even have summarized them all or understand them all. And someone from Spiegel, which is a big German newspaper, called it the first post-factual war. And this is kind of an interesting term I think. And if you have your opinion already, basically no matter what opinion you have, you can find it on the internet and you can find facts that support your opinion. And this kind of spread is even sped up by social networks. Today, really all you need is a phone with a camera and Twitter and you can be your own producer of news. And this didn't really exist before and kind of makes sure that a lot of these reality filters are not there anymore. So traditionally, traditional media really doesn't tell you what to think about a topic or how to think about a topic, but instead they tell you what topics to think about and this filter is kind of lost now. Obviously, I have seen that there's a huge diversification of sources of news. Now I've seen that movies usually used to belong to someone very specifically and it used to be something rare. And you used to know where it came from and that's not true anymore. I myself work in movies, so this is what I want to look at. So what can be the motivation of people falsifying movies? First of all, just for entertainment. I have an example for that, the typical example for that is actually Z-Hire. Of course, there's also manipulated pictures and videos and they were not made to hurt someone but to entertain. Then there's political influence or propaganda that can be made from states or private people or even parties. I remember a nice picture from Iran, from Iranian military. They were not very happy with their rocket launches. So someone just opened Photoshop and basically had the picture that showed all the rockets launching even though some of the launches failed. Now profit is the thing that we see in the American election. Fake news papers that apparently had websites and these Macedonian teenagers found that they could actually make money quite easily. So in Germany this is not really the problem right now and can't be compared to the US election. But we have to decide where to draw the line. It was very extreme examples that we now saw from the US and there's a lot of lies in this. But what about clickbait? So things are getting made very important and also traditional media are jumping the train. But it's not behind the scenes something else is happening. And also attention. So if 20,000 accounts from Twitter are retweeting one count and you get a lot of retweets on this, people are very happy about this. The people that are tweeting this. So it's fun and attention that the people get that posters on Twitter. So let's look at the different shorts. There's the satire. So somebody that understands that for him it's not really fake news or it's not really lies. And let's look at one example from the onion. It's a big, it's like comparable to the German newspaper, the postulion or the online newspaper. They published an article in 2012 and they copied it one to one from Iranian press agency. And the title was white people living on the countryside prefer Ahmadine Resa to going to having a beer or going to a baseball game. And for the people that didn't get that satire, it might also be wrong news, but it was meant funny, not real. So a lot of times also there's old pictures and they're put in a different, in a wrong content. And it's the simplest way of doing something. So the pictures are not wrong. You don't have to Photoshop the pictures. It's real pictures. You use Google, you find some pictures and you put the wrong context from title below it. Then the next step would be that you would have pictures that somebody works on with Photoshop or something like this. I remember there's one picture from Lebanon showing Beirut and there's been a lot of smoke because it was bombed. And the smoke has been, they put more smokes on it and that's not right of doing it. It shouldn't be part of the news. And last of all, there's stuff that is, there's no real content or no real at all. It's all completely fake stuff and you might have actors doing something like this. And we also looked at these stuff and tried to find out if it's right or not. And we found that it's a viral network that made money with this, with something that has been viral and a lot of people looked at it. So since about two and a half years I'm in this part of verifying this news and seeing if they're fake or not. And it all started with in Munich where some people were shot by terrorists. And we saw that there was one single event, there was a lot of different fake news on this and let's look at some examples. And the big German television program also showed that picture. A student has posted a picture. So it's right that far. But this is not a picture that has anything to do with the event that went on in Munich. So if you search for that picture you found that it's an event from a year ago and very simply by using the reverse search on Google you could find out that this picture is a different context that was taken on. Twitter account true news if you can believe it or not. And they ask is it real? Is it really an active shooter or maybe you decide yourself if it's a child or not. And you see if you really try to find out where this picture comes from it's a different picture again. It's from Manchester where the police made a test and there was some actors playing a role. It's a nice picture of something that was manipulated and again it's the right place. It's the place where it took happening the shooting. But the person on the picture was put in. And Sam Hyde that's somebody that's a comedian and he was put in with Photoshop like this and he's famous for putting in relation to these such events. So what can we do? First of all don't believe anything. You have to be very careful if pictures look too good to be true. Keep your distance on cases like this. And it's very true and it's very short and after the event took place there might be a high probability that it's not as true as it seems to be. Turn on your brain, think about if this could be true or not. And look at details, details, details. Might be something in the background if you see something or there's some advertisement to be seen on the picture. There was a video from Chemnitz where people said there was not a hunt on people and it was video footage from a year ago. But you could see an ad-word in the back of the video and there was a theater play to be seen on the background and the date was a recent date so you knew it was a recent video. So if you open your eyes and look at it you can really see what's going on and you don't need to judge on it but you can just look and see what you see. And if you compare that with what it should tell you you might see that some things might not fit together. So the next picture should come from Thailand. You see youth who was locked in the cave and it was a video. You see people putting stones out of the way so they can get out. But what kind of stones is it? They're quite round and the cave and it seems to be in the middle of a mountain. Why should there be round pebbles, round stones over there? So what you actually see is divers who do diving in the cave and they test something but it might be 10,000 views that people looked at this. So if you look at it can you see something where you can recognize a certain place? Is there special buildings on it? Are there like roads that you might remind you of something? Or there are some rocks in the background that might zoom in? Does the weather fit what it should be? So if it's supposed to take place in the winter and everybody runs around in the t-shirt it might be a hint that something is wrong with this. And this you can get a weather forecast or you can also get a historical weather forecast and see what the weather was like on that specific day. Was there rain? You can look at the number plates, you can look at the signs on the streets and you can see if the shadow that people and it fits together with the time that this is supposed to take place. So there's some links you can find on the agenda where you can also read all these details and look at all the tools and there's links to all the tools that are used and there's a verification handbook that really helps you doing the stuff that I'm doing here. The next step would be to find the primary source of the picture. So if there was a catastrophe you know so if there's one video you know a lot of people would download it, they upload it again you can find it in a lot of different places with the new media. So it's important to find where this really originated, where it really came from to find out if it's true or not. We have one video that was supposed to show an earthquake on a Greece island and we looked at the uploader and we saw that he posted on his account a lot of pictures of that specific region just before so it was very likely that he was really there. You can use a reverse insert not only over Google but also over different platforms through a lot of different search engines you can really see who's the uploader, who's that person and you can also try to find out where the scene really took place. So if you really don't find a primary source, you might find out how many explosions that have been in Munich on a specific place. And you have to think locally. Local specialties in the city. How does the uniforms look? The people from the police, how do the policemen look? How do the firemen look? How do the license plate of the cars look? Is there local experts? So if in Germany and there's a small city there might be a way you find out somebody who's working in that city and who might know about more about this and maybe have historical data about the city. The tools that we use and we know a lot of times are not the best tools to use. The next video we're going to look at is going to be an explosion in front of a... You can see a lot, you can see there's smoke, there's some people, there's some power lines going over the street and that's the scenery. And it should be in front of the US Embassy. So I checked, you know, let's look at Google Maps. Where's the embassy actually at? Let's look at the satellite picture. Looks a bit strange. The streets look very different. The don't pictures match together. So maybe Google wasn't able to update fast enough. I also checked Bing, there was also some wrong pictures of this. And there was an offset on the maps so the locations didn't fit with the aerial pictures or the satellite pictures you could see in there. But there's also the Chinese Google. And there the maps were much better and there was no offset on this. And they also have the street view on this. And there you can see there's a sign which we saw in the video. There's the building that we saw with the spike on top. There's the power lines go over this. So you can be pretty sure that the video was taken on the place that they actually said the picture was taken. So don't rely on the first source you get. Be critical. It's not only Google, especially in different areas of the world there's different people who might have better information than this. This video shows the tsunami on Sulawesi a couple of months ago. And here we see a mosque that looks pretty badly beaten up and a colorful wall. And I cut some of it and you can see that it was taken, the video was taken in a colorful room. But it wasn't said where exactly. But also on Sulawesi this coast is actually quite long and to look at Google Maps seemed like it would take a long time. But instead luckily there's programs like Wikimapia. They want to categorize everything. So then you can look at what coast it is and then you can choose a category like mosque. They also have category like churches, memorials, schools, government buildings, which we used once for a video from Syria. That was ostensibly taken between a school and a government building. And you can look at both of these layers and see whether that actually helps. And this is much more comfortable than using Google Maps and searching for house roofs. Why should everyone be able to do that? Well, we want to create a kind of anti-publicity. So these videos, they're shared so much. So everyone who can should be able to basically tell everyone what the truth is. Now what I've shown you was actually not very severe, but even there we should make sure that it's not shared as much. And these fakes can also make kind of false narratives. Now what I've shown you wasn't very important for public opinion, but we had pictures from Munich that showed completely different places where the crime took place. And the murderers who weren't really the murderers, and if they're distributed and shared on the network, then they might actually kill themselves or might be killed even. And in the worst case it can actually influence societal debates. There's a special to present a picture from a crime. So how do we do that? What kind of tools exist? I have an overview. The links are always there from Syria. You can use the website there. Almost all of the rebels use their logos in their pictures and videos. I can't read Arabic. So I can't really see what kind of group is behind this video. And with this website I can just upload my videos. They extract the logo and tell you what kind of group this is. And also show you the Wikipedia. This is also logos from TV networks. So if you don't know which network has broadcast something and you want to find that out, this is a very good source. The details in the background. Who knows how Iranian, how Iraqi plates looked in the 1950s. These are also for all African countries. So if you think that a video was taken in a certain country during a certain time and you can kind of see the plates, you can identify it. The place of the sun you can calculate with this tool for example. If you put in the place and the time, it will tell you where the sun should be. So it will tell you whether the shadows seem right. So this shows basically at this time an object with one meter size gives a shadow of 1.71 meters. Now you can roughly check whether a video makes sense, whether it makes sense that it was taken at a specific time. Now even Facebook accounts give you some ideas. Even if you look at it as someone who's not of a friend. So what is not private but is public is what you liked or what you commented on. And this is always public and some websites can kind of find that out. So you can look for photos or posts. You can look for places that someone looked for. You can look for relationships between people. You can look for when one person mentioned the other person. These are things that you don't see with Facebook in the normal, but instead you see it here. Even if you told Facebook that you want to complete privacy, you can still use that. Now for English speaking accounts, this kind of search actually exists in Facebook. And you can also use users who like. But there's still some sort of UI problems, issues with using Facebook itself. This is why there's good websites that actually do that for you. Then there's a plugin that I recommend for everyone, Invit, which offers many functions like getting keyframes from videos showing when a video was uploaded first. Data upload data and the time is not really nice because a lot of the platforms have a lot of differences depending on where you are and whether you're locked in. And this Invit project offers a standard that is comparable. It offers a magnifier. It offers for you to do a reverse search and to look for the mirrored image as well in a reverse search. Because some people think that you won't be able to find it using a reverse search if they mirror the picture. But this is actually a pretty amazing tool for you to use. Now again, verification is no rocket science. But it is necessary for you to be on time, to kind of make an effort and to train. There's a lot of programs, there's websites, there's plugins, there's lots of documentation on how to do that. So I'd be happy if I awaken some interest in some of you. If you need some practice, please follow QuizTime, which is an account by some journalists in German, some German journalists that basically share quizzes, share practice for you and they show, they allow you to practice yourself, but they also show how to do that. Ah, perfect. Now this is almost the end. We still have some time, so if you have more questions, please ask them. So we have about ten minutes for questions and we start as always with a question from the Internet. Is there some sources where you say you have to be careful in general if there's some news coming out of that source? Like special newspapers? It's difficult. I'd say in general the Internet. I think you should always be critical. What we have said in our work is even if just because some other media reported it, it doesn't mean that they could not have been tricked. So just following other media is not enough because everyone makes mistakes. So just taking it from other media is not an argument for the various vocation, for the various cities. There's a content on taking fake news out and also verifying this fake news. And one of the problems is that a lot of times it's more work than verifying the news and finding out if they're right or not, then just retweeting it or repeating it. I don't think there's anything you can do about it. So as a journalist, I think you're a filter. So basically you can't say, well, I want to falsify each and every lie on the Internet, but instead if you write about a topic, then you can choose one picture and you can make sure that this picture is actually right. I don't think anyone has the capacity to make sure that there's no more lies on the Internet. How trustworthy do you think are the tools you're using and do you try to use even more? I think they're good in what they can do, but there's always some interpretation and experience necessary and oftentimes it's really just thinking hard in your head. What I don't really like are photo forensic tools that work with ILA that try to show you whether something was added, but there's actually no scientific basis for that. You actually see the photo and this analysis, but it's basically your interpretation. Is this fake or is this not fake? Because sometimes it shows that it's fake even though it's just bad lighting. Other programs where you still have to do something yourself, I actually trust them, but those forensic software I don't really trust. You didn't mention a lot how you find the primary source of a video or some news. So the first step would be a reverse search because most of the fakes that you see are kind of recycled material. So if you do a reverse search, you can find the original video if it was five years ago and then if it was five years ago, you don't really care about who actually made it five years ago most of the time. Otherwise, it helps to just look at what does it show and then just search for those keywords. We have one video from the Egypt Spring that was supposed to show people demonstrating kind of throwing a police car from a bridge. In reality, it was actually just someone put in reverse by accident and drove from the bridge. So this is a way of this where you could search by just using keywords. Thanks for the good presentation. Come a little bit closer. Now my question. Reuters, banned all raw material, they only want JPEGs. So look at your example with the missiles. Do you think why the reason for this is why a source that's really hard to fake would be rejected but only the easier to fake one? Honestly, I have no idea and I can't really explain it. I didn't even know this before you told me. Thank you very much for my side. People talked a lot about these deep fake stories. Did you see a real deep fake in real life, not the examples that were on the television? No, I haven't seen it in real life before. I've never seen it that was sort of relevant enough. There's a couple of things that the future might bring but for now we can still distinguish deep fakes by basically looking at the details because for deep fakes a lot of the time the background kind of shakes around a little bit. It's not as sharp, not as in focus. The software that was supposed to be able to fake voices, we haven't heard anything from that. I think we will not hear anything from that either. That's one question from the internet. Do you have one example that was used in school circumstances where pupils were treated with fake news? No, of course that would be great but I think it might take a lot of time until schools actually teach this and this might actually be something that has to be in sort of the larger topic of media competency. One question, thank you very much for your presentation. It was really good, thank you. Everything you can do with algorithms. So with your Facebook friends what perspective is this something that's important for you because it's also a kind of fake news that they produce in a certain kind of interest and it's really hard to find out what algorithms did why put something under them. Is that something for you? I'm not quite sure what you mean. Do you mean like the way Facebook structures what you can see? Yes, that's what I mean. Yeah, that's very problematic of course. That's also one of the reasons why those fake news are shared as much. There's echo chambers that have the social networks that try to present content that's interesting for you. Do you have a tip how to make this more transparent? Well, I don't know how to make social media more transparent. Of course we need to find this but to some degree it's everyone's decision whether they want to use social media to consume news or whether they want to use traditional news. There's different people who want to classify websites. So it's a satire website or the left-wing or right-wing website. And to also show it to the user so if you surf on a website you would see what kind of type of website it is. That's a very difficult topic as well. With Facebook I'm not quite sure. Well, if there's many people that have classified a link as fake or if some experts classified it as fake then it should at least be put on to this post. I think this is a better system than having websites actually be classified. Also something like satire lives of being able to fake something a little bit even though some people don't understand. Thank you very much for the presentation. A question maybe a suggestion. So before you do the research every media does his own research but if you find out that something is really a fake do you also forward that information to other media and publish that information so everybody knows it's a fake. So the way I know it from other media is that we have conferences where we kind of share stuff and we go into larger groups and we talk about these issues. I know that in larger media with more than one product they tend to have only one verification team or they share their info. The last question from the internet and one of the talks of the last year is somebody had the thesis that fake media is not as big as it was suggested. So the question to you is do you think that this has changed in the last year. Is there more fake media is there no less fake media. I wouldn't necessarily say that it became different. The patterns are the same. The motivation is the same. People that want attention. People that want to convince people what has not changed that all of that has not changed a lot. What we've seen is that there was a lot less breaking news about fake news. So 2016 was a pretty bad year in that way. And especially if it's something that incites emotions like terror acts in Europe or something. The output in fake news is a lot larger than it usually is. Whereas in normal circumstances we have less. And in 2018 it was actually better. Now thank you.