February 26, 2024 • 47 mins
Prepare for an enlightening journey through episode 93 of the Criminology Academy podcast. Explore the dark side of the internet with cybercrime expert Dr. Eden Kamar, who delves into the disturbing trend of child online grooming. Through a detailed discussion about her groundbreaking research and studies, we also explore the role of parental guardianship in mitigating online grooming, the techniques employed by predators, and protective measures for children.
Unveil the innovative use of chatbots and artificial intelligence for data collection and analysis of online sexual grooming techniques, highlighting the potential in safeguarding children and prosecuting offenders. Learn about the alarming findings, including the observed trend of more sexually knowledgeable children being specifically targeted for offline contact. Reflect on the importance of understanding the grooming process's non-linear nature, which can often be abrupt and direct.
Dig deeper as the conversation explores the issue of webcam exploitation, particularly by online groomers. Recognize the deception within certain platforms that enable groomers to manipulate the victims' cameras without their awareness primarily through 'Whereby' as suggested by Dr. Kamar's study. Raise pressing questions on the safety design of such platforms and the role of software companies in securing user safety.
Tune into this insightful conversation and empower yourself with knowledge against these cyber threats.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
Music.
(00:14):
Hi everyone, welcome back to the Criminology Academy podcast where we are criminally academic.
My name is Jen Tosley. And my name is José Sánchez.
Today we have episode 93 and for this episode we have on Dr.
Eden Kumar who will speak with us about cybercrime, specifically the online grooming of children.
Dr. Eden Kumar is a postgraduate in the Evidence-Based Cybersecurity Research
(00:39):
Group at Georgia State University.
She received her PhD in Criminology specializing in in Cybercrime and Cybersecurity
at the Hebrew University of Jerusalem and her MPhil in Chronological Research
from the University of Cambridge.
Her research focuses on cybercrime and cybersecurity, namely technology,
facilitated sexual abuse of minors, phishing attacks, illicit markets,
(01:01):
malicious codes, and information system vulnerabilities.
In this episode, our conversation centers around two articles and one op-ed.
The articles, The first one is Parental Guardianship and Online Sexual Grooming
of Teenagers, a Honeypot Experiment Published in Computers and Human Behavior in 2022.
(01:22):
And the second article is The Relevance of Targets, Sexual Knowledge and the
Progression of Online Sexual Grooming Events, Findings from an Online Field
Experiment Published in Justice Quarterly in 2023.
Both articles were co-authored with David Myman, David Weisberg, and Diko Shaba.
(01:42):
We're also going to discuss op-ed in the conversation co-authored with C.
Jordan Howell titled Online Predators Target Children's Webcams Studies Fine, published in 2023.
So with that being said, let's bring Eden in. Hi, Eden.
Thank you so much for joining me and Jose today. We are excited to speak with you about your work.
(02:04):
Thank you for having me today and for inviting me me to be on the podcast.
It's great and a pleasure to be here with you guys. All right. So let's get started.
We're going to go over some background first, kind of get the lay of the land
when it comes to your research area.
And so we have had a couple of episodes in the past talking about cybercrime,
(02:24):
specifically Zoom bombing and the dark web.
And so in this episode, we're going to be speaking about something related but different,
which is online grooming. And so we want to just set the stage and begin with
a definitional question, which is our specialty here on this podcast.
And so what is online grooming? So it's very difficult to define it because
(02:48):
there's many different definitions and common ones that are used often refer to offline grooming,
which refers to grooming as any manipulation.
Manipulative behavior towards children and the child's environment to convince
them to do any sexual behavior.
Now, this changes when the behavior moves to the online environment because
(03:11):
the parents or the environment of the child is kind of different.
They're not around there. They don't know. They don't see. They're not aware
of what's going on usually.
So what What happens is the entire manipulation is toward the child.
And so the definition that I adopted, which is by Joblay et al.
(03:31):
2021, she defines it as any interaction between an adult and a minor with the
intention of sexual interaction, either online or offline.
So it's more about the interaction between the adult and the minor rather than
the environment outside of this interaction that influences sex.
Whether sexual offense occurs or not.
(03:53):
So that's how I adopted it and how I see online grooming.
And I think it's an inclusive one because it takes into consideration the shift
between the offline and the online, which previously has been disregarded.
Yeah. And I imagine that this is a difficult thing to study and get estimates
about, but do we know what the prevalence of online line grooming is?
(04:16):
Like how often does this occur?
So there is estimates based on police reports, which has criminologists we know
as the tip of the iceberg.
And in girls, it's one in three, according to the police records.
In boys, it's I think one in nine, if I quoted correctly.
(04:37):
But we don't really know. And we only look at female and male whilst we know
there are other there are genders out there and we don't know the reports towards those genders.
Research that looks around online grooming up until 2019 said that there was
a decline in online grooming.
But since COVID-19, we see a significant increase in the number of cases.
(05:02):
And we also see differences from before to after and more use and manipulation
of technology to commit sexual offenses against minors?
One in three and one in nine. That seems really high already,
let alone if it's the tip of the iceberg, like we typically know police reports are.
Yeah, that's pretty high up there. So there's like a few things that we know
(05:25):
about crime and victimization.
And one of those is that usually for most crimes, it'll involve someone you
know, rather than like a complete stranger.
When it When it comes to online grooming, does that typically hold or who is
most likely to be doing the grooming and what does the usual grooming victim look like?
(05:47):
So in comparison to offline where the victim and the offender have some relationship,
they know each other, either they've been in contact somehow,
either it's a family friend or a coach, sports coach or a teacher,
or someone related from the environment of the minor online,
(06:08):
it's a complete stranger.
That person can be in Australia, in the UK, India, they can be anywhere and
they can be talking to a minor in the US or any other country around the world.
So the difference is not only that it's a complete stranger,
it's also someone that can be miles and miles away and they do not necessarily
(06:33):
have to leave the comfort of their home to commit the crime.
They can just stay in front of their computer.
And achieve whatever goal they intend to achieve at that point.
In like an anonymous fashion too, right? I'm assuming most of the perpetrators would be anonymous.
(06:55):
Yes, and also it's hard to know how much of the information they provide is
true or not true because they want to disguise their identity at least until
they do some risk assessment of who you are and how much they can trust you. All right.
So when Jose and I were talking about online grooming, my mind went to shows
(07:16):
like Criminal Minds or other popular TV shows that have had some element of
grooming, whether it's in person or online.
And so people who are familiar with these shows may imagine someone looking
to groom a child or a teen goes to a chat room and catfishes someone into an
online relationship by pretending to be someone around the victim's age.
(07:40):
Is this actually how online grooming works or how is contact made between the
offender and the victim?
So chat rooms are still quite popular for this type of behavior,
although there are some use of some other platforms nowadays like Facebook, Instagram.
Dating apps for teenagers are also quite common for groomers,
(08:03):
but chat rooms are quite popular.
And the offenders, some of them are still using deceitful behavior and deceiving
behavior and pretending to be minors.
The majority of them do not feel the necessity to do so.
And they just state that they are old or just say, I'm old or refer to themselves
(08:25):
as daddy, which is a nasty way to describe themselves. myself.
So I think that that's going to set us up for the first paper that we're going
to discuss, which is titled Parental Guardianship and Online Sexual Grooming
of Teenagers, a Honeypot Experiment.
And so our first question about this paper is, what was the motivation behind it?
(08:46):
And so what were the gaps that you're trying to address with this paper?
Many studies looked at guardianship in the context of online grooming.
The majority of them used surveys to collect data.
They referred to where in the house the computer was based, like whether how
much involved the parents were, if there was parental conflict in the house
(09:08):
that the child was disposed to.
They looked at different stuff, but I was more curious about how the way that
children are navigating or like using data.
The online environment, and the context of the household that affect the context of guardianship.
So it takes us a little bit to Rosgood's theory when he talks about unstructured socializing,
(09:34):
when he says that when minors are spending time outside without parental supervision,
they're more likely to conduct,
like to be criminal, to do criminal activity.
But they are also more likely to be victims of crimes because there's no one
there to supervise them.
And so relying on that with the context of guardianship in general,
(09:58):
it was interesting for me to see like how the parental supervision of children
who are going online, because the online environment nowadays is like the playground of the 90s.
Children, instead of going outside to play in the playground.
They go online and they play games and they socialize and they meet people.
(10:19):
So the way that we socialize changed.
So we also need to adapt how we look at theories that applied there back then
to how they can apply nowadays to criminality and victimization.
And so I looked at how we can relate the theory into the cyberspace.
And in this context, parental supervision would be a parent sitting next to
(10:42):
the child when they're spending time online.
And that's what I termed active supervision, when parent is actively supervising
their children's online behavior. behavior.
Now, although Osgood didn't refer to it, but there's another theory that I drew
on from Ryan's that he looked at different levels of guardianship in the context
(11:03):
of neighborhoods, and he referred to active, proactive,
and passive guardianship, where passive for him was when someone was just in the house.
Someone was in the house and that prevented someone to burglar the house.
So I wanted to see that in the context of online grooming, the mere presence
of a parent in the house might affect how the groomer reacts.
(11:25):
And how he changes his behavior towards the entire situation.
So what we did is we looked at those on passive or active parental supervision
compared to no supervision at all, which is the most common nowadays that children
are just playing online,
to see how once the offender encounters a different situation,
(11:48):
how he reacts to the situation, and whether it is effective in terms of guardianship.
Because asking parents on a survey or children on a survey how their parents
behave and then refer to it if it affects,
you know, guardianship is not as effective as if we actually talk to the groomers,
actively engage with them, and we see how they react to different scenarios
(12:12):
that we can actually measure the effect of it on the progression of the grooming process.
And so that's kind of how I came up. And it's really interesting,
these different levels of parental supervision and how that impacts these different
interactions and potential grooming situations.
(12:33):
As I mentioned the title of this paper, you mentioned that this is a honeypot experiment.
Can you tell us what a honeypot is? Yeah, so honeypots is an environment to
which we lure an offender to be able to study his behavior, his reaction to different scenarios.
(12:54):
It's kind of like a mousetrap, right? We pretend to be, in the context of this
study, we pretended to be 13-year-old girls.
We created fake users of 13, 14-year-old girls on a sample of 23 chat rooms.
And then we just waited, really, for them to approach and start communicating with us.
(13:15):
And once they started engaging in communication, to ensure that we are actually
talking to offenders and not talking to minors, which is also a bit problematic,
and I'll talk about it further later.
The first question was ASL, which is age, sex, location.
We wanted to make sure that we're talking to adults and not minors.
But then it's also a limitation.
(13:36):
We started talking in the beginning where some of them are using a deceiving
approach of pretending to be be minors.
So we might exclude those who use this deceiving behavior from this study,
but we had to ensure that we are only talking to adults or to individuals who
identify themselves as adults.
(13:58):
And that's why we asked this question initially.
And after we made sure, based on the treatment that was assigned to the user
that the offender approached to, then the conversation would continue based on that.
I've heard of honeypots before in our other episodes, but yours was a little
(14:18):
bit different where you used a honeypot chatbot, which was something new that
I hadn't heard of before.
Can you tell us more about how you used this honeypot chatbot and what about
collecting the 639 communications you therefore used in this study?
Of course. So a chatbot is basically automation of a conversation flow using
(14:42):
natural language processing.
So basically, we have a script. We try to predict any questions that the groomers would ask us.
And then based on this, we created a script or flow of conversation with a set
of answers that we wanted the machine to reply.
And then each treatment had a different conversation flow.
(15:05):
But since it's the first study ever using chatbot in the context of online grooming,
it was very tricky to predict any potential behavior that the groomer would
or any potential question that the groomer would ask us.
So what we did is the bot was functioning in a way that if it wasn't 100% sure
in the question that was asked, it would email me as a person who is supervising the bot.
(15:32):
I can read the email with the conversation flow.
And then, and it will say like the email would be like, the title would be active
or passive or no guardianship.
So I know exactly to which treatment it belongs.
And then I would have the conversation flow. I can read through,
then I can reply through the email, and then the email would be sent automatically,
(15:54):
like the message that I'm sending an email would be automatically appearing
in the chat room and sent to the offender.
And that's how we collected it. At the end of each data collection,
the bot automatically saves everything to an encrypted platform to ensure that
the data is secure and also no one else can access it.
(16:16):
And we anonymize information that is not supposed to appear, any identifiers, and
So that's how it works. And it has also the ability to flag us.
So when we designed the bot, I wanted to be able to know which conversations
had keywords that I know are groomers' words, usually.
(16:40):
So it will flag this conversation.
So it would just basically drop me an email saying, red flag on a conversation.
So I can go to the chat room or like go to the machine where the bot is running
and read through when bot is still communicating.
The bot was running through. So we had in total, we were sitting on 23 chat rooms.
(17:03):
We had three users on each and we were running it through 14 virtual machines.
So we needed a lot of power and we
had about three emails in total
communicating communicating with me and the
bot communicating with itself so it was a
huge operation to make this all work but
(17:24):
it's a cool way not only to collect data
but also to save time because if
I had to do it without the chatbot I
say it always is a joke but I would retire before I had
a piece because it was so
much time consuming to communicate with each one
of them individually So having the machine answering automatically based on
(17:47):
those scripts and me just supervising whenever and assisting the bot whenever
it's not sure allowed me to collect 639 conversations in about three and a half months,
which without it would take me probably a couple of years.
Yeah, no kidding. That's really cool. I imagine it still took you quite a bit
(18:10):
of time considering this is the first time, like you said, this chatbot was used.
Did you have to intervene like when it was having trouble figuring out what
to say a lot or was it pretty on top of the conversations? conversations?
Unfortunately, yes, particularly in the control group.
So in the no guardianship at all, because the conversation went on and on and on and on.
(18:34):
And it escalated to points where I wasn't expecting it to escalate.
And so I had to intervene because I didn't foresee this going on so long.
So that's mainly when I had to intervene quite a bit.
And that was also challenging for me those conversations
that went on and on because it escalated to
(18:57):
points where it got very nasty but just
put it like I can imagine but I don't want to yeah so just to break this down
and be more specific so you had your two treatment groups then in one control
group was the only thing that differed between these the active passive and
no parental supervision Or were there other things too?
(19:19):
Just this. I mean, they had different... So the way we designed it was that
I went to this website. I don't know if...
Anyone mentioned it before, there's this website that calls this person doesn't exist.
So it's an AI that develop images of, it takes a lot of images from online and
merge them into one image of a person who doesn't exist.
(19:41):
And so I went on this website and I downloaded images of people who appeared to be 13 year old girls.
And I use those images instead of, you know, real people. And then we had random usernames.
So literally just Googled girls' usernames to chat rooms and ideas.
(20:04):
And then I had a list of whatever people previously used.
So I just scrapped and used those just to make it look like a 13-year-old girl's chat room profile.
And then what we manipulated was the answers.
But then that was tricky as well. So initially, I thought that based on all
(20:25):
the literature and what we know about grooming, that offenders are doing risk
assessment and they ask, hey, your parents are around. Are you alone? Are you in your room?
Where are you? Something, right? So out of the 639 conversations,
we found that about, if I remember correctly, about 57 conversations,
only in 57 conversations, they actually asked.
(20:47):
Class and it was shocking because I
thought that's going to be one of the most prominent questions
so in the first few days I had to kind
of change the design because we designed it so okay whenever
they ask we reply but then when we notice they're not asking I was like okay
either the literature something in the literature is not as it was supposed
(21:09):
to be or something is off and then we were like okay so how are we going to
move with it so we decided that whenever Whenever they started asking sexual questions.
Questions, that's when we're going to introduce our treatment.
So when they ask, for example, something that is not too nasty,
(21:29):
they would ask, have you ever kissed?
Or what are you wearing? If it's the active treatment, the bot would trigger, I'm not alone.
My mom is next to me. I'm not allowed to be on the computer on my own.
And if it was the passive one,
it would be, be right back my mom is
calling me from downstairs so they would know the mother
(21:52):
is in the house and then we would wait five minutes and then we message back
back and see like if they
pick up or they drop it and we hypothesize
that the active parental guardianship would have more impact and would be more
likely to reduce the grooming process from continuing and even when we compare
(22:15):
compared to the passive guardianship.
And the results indicated that exactly.
We found that 92% of the conversations, so the likelihood was reduced in 92%
compared to the control group.
And then we had 57% reduction for the passive.
(22:38):
And then there was 87% between the active and the passive. So it was the active
parental supervision was substantially reducing the likelihood that the grooming
process would continue.
But then something that wasn't on the paper, those who did continue the grooming
process shifted the abuse from the minor to both the minor and the parent.
(23:05):
So they weren't just satisfied with abusing the minor. They were like,
okay, so does your mom want to take part of it as well? Wow.
Yeah. Or questions like, is your mom old as you? Okay. Interesting. Interesting.
I wouldn't have expected that. Yeah. So, you know, it goes back to the literature
that talks about the fact that some sex offenders are just sex offenders in
(23:28):
general. It's not just against minors.
It's also against, you know, they perpetrate against adults as well.
They do not differentiate.
They don't have preference necessarily.
So they just combined the two in this case and shifted to both the parent and the minor.
Well, parent. the big parent in this situation.
(23:52):
But it was a bit surprising to me to actually see it.
I think the first few times I was very surprised that's the case.
Do you know how prevalent that was? How often were they like,
bring your mom into it too?
I never quantified it, to be honest.
(24:13):
I don't want to give them numbers because I don't know the exact numbers,
But it wasn't uncommon for those who weren't deterred by the guardianship.
Right. It was enough to where it was kind of a pattern, just like a one-off.
Yeah. All right. So just to quickly recap. So you found that having the active
guardianship really reduced the risk for grooming.
(24:37):
And then that was followed by the passive guardianship. And then last was no
guardianship. It was like in comparison. The first model was in comparison.
So we compared active with no guardianship, passive with no guardianship,
and then we compared the active to passive.
And in all cases, in both two models, the active and the passive was more likely
(25:01):
to reduce the online grooming than the control group.
And then when we compared the two, it was the active that was more effective
in reducing the likelihood of grooming from proceeding.
All right. So you sent us another paper too.
And so this next one that we're going to talk about is titled,
(25:21):
The Relevance of Target Sexual Knowledge in the Progression of Online Sexual
Grooming Events, Findings from an Online Field Experiment. So same question we asked you last time.
What was your motivation or why did you decide to study this and write this paper?
So there is this assumption that naive children are more likely to be victimized.
(25:44):
And I read a few papers that talk about how sexually knowledgeable offenders
that were interviewed says that they actually look for those who are sexually
knowledgeable compared to the naive.
And I decided to test it. I mean, let's, why not?
I mean, we need to make sure that we differentiate between sexual knowledge.
(26:09):
And sexual experience. It's two different things. Sexual knowledge is me knowing
certain terminologies, understanding what it means.
Sexual experience, it means you have tried it, have done it before.
We don't test this. We test only the knowledge.
So whenever there was a situation, if we were asked if we tried something, it will be no.
(26:31):
But whether we know or not know, No, it's a completely different thing.
And I think this is a very relevant thing for us to consider.
We want to develop awareness trainings, right?
We want to prevent and we want to make the children aware of the risks.
And one of the things we need to consider is what we're teaching.
(26:51):
Are we teaching them to avoid to talk about certain things because it's riskier?
Or actually, if they talk about it, it might not have any effect.
So for us to exactly know how to develop trainings, we need to know that.
So that's when I kind of decided to study the effect of sexual knowledge on the grooming process.
(27:14):
And what we did is we used the same approach as in the first experiment.
But this time, we only had two users on a sample of, I think,
21 chat rooms, if I remember correctly.
And we manipulated whether the child knew what certain words mean or they did
not know what it means and see how the groomers react to this particular scenario.
(27:41):
And in this case, we found that actually, compared to what everyone's saying
about naive children being more at risk of grooming,
the sexually knowledgeable children were much more at risk of grooming and were
more likely to receive requests for offline contact,
which is much more alarming than the actual previous findings because of the
(28:06):
physical and psychological impact that offline sexual intercourse can have.
Sexual abuse has, not minimizing the online one, because online one is substantial
as well, particularly with reliving the victimization every time an image is
shared or reposted or whatever.
Both of them are bad equally, but I just think that physical abuse is taking
(28:31):
it a bit step further. And so...
That was very important, I think, finding for us to then be able to not only
think about like awareness trainings, but also think about sting operations,
because a lot of the times sting operations, police sting operations are like
(28:53):
it gets to the point where it gets to court, but then the court dismiss it with
the accusation that the conversation was leading.
Officer was leading the groomer, and therefore the evidence is dismissed.
And so we demonstrate with this study that just by saying that we understand
certain terminologies, sexual terminologies, without even having any sexual
(29:18):
experience, still lead to an offline encounter. counter.
So if sting operator officers just use this kind of approach,
they might be able to prosecute the offender without having evidence dismissed
later on when it gets to court.
So I think it has substantial contribution for both areas.
(29:43):
Yeah, absolutely. I want to go back to this point that you said about this request
for an an offline meeting and a point that you made earlier where online grooming
the offender and the victim might be,
you know, miles, countries apart.
Did that play any role into this request for an offline meeting?
(30:05):
Or maybe you haven't looked at this at all. And I'm just hitting you with a random question.
But I think it's interesting, this offline meeting request, when potentially
one person could be in Australia and the other one in Canada.
So I haven't looked at it to answer to your questions.
I haven't looked at it. To one of you that I do remember,
(30:29):
I remember the chatbot was in New York and the offender was saying that he was
in the East Coast as well in one of the conversations that I do remember.
But I haven't looked at it in general to provide more.
But it's not uncommon for online groomers to travel.
It is quite common. And for example, in the UK, one of the ways that the National
(30:55):
Crime Agency managed to prosecute online groomers is when they land in the UK.
So they communicate with them to the point where they arrange offline meeting.
And once they land in the UK territory, in the airport, they arrest them in the airport.
So it's not uncommon for offender to be in trouble.
(31:17):
Interesting. So, one of the things that you cover in the paper is like the process
of grooming or like these various phases of online grooming that groomers tend to go through.
How were you trying to identify these phases or like flagging like,
okay, we've hit like a...
(31:38):
I should have written down the names. Like the report building,
sexual content assessment, those phases, were you keeping an eye out for trying
to track where you were? So there's different types of like grooming process, right?
Typologies there is like there
isn't really one that is agreed on
(31:58):
i mean there is a connell's one that it's the
most cited one that talks about report building and
then moving on from one case to another in a
linear way but from my studies i've noticed
that it's not as linear as people think it is it's not like the groomer start
a conversation with building report they might just go straight into what they're
(32:20):
after and like send you a link to a video call and ask you to jump on the call
and let's just do whatever day he wants to do on the video call.
Or he immediately start with saying sexual things without building any initial report.
But then when, for example, we always had to ask, no matter what is the first
(32:44):
question that the groomer asked or the first message that the groomer asked,
we had to kind of like take it a step back to ask about the age.
So it would be like, okay, now after I already asked, she asked me about the age.
So now I need to, maybe I should start building a bit of rapport with her before I move on.
Or they might just answer the question and then immediately send the link again,
(33:06):
like just come to the video, like to the video call, just come to the video call.
So it's not linear or progressing in a way that we can describe.
What we do find is that something that is not really being discussed in previous
literature is that there are other pathways into grooming that do not appear in the literature,
(33:28):
like the use of URLs that lead us to websites that can be utilized to different things.
Okay, so I think that can get us into the last piece that you sent us that we're going to discuss today.
So in May of 2023, you wrote a piece for The Conversation about online groomers and using webcams.
(33:50):
And the data that you presented in this article is part of a study that's currently under review.
Can you give us an overview of this study and sort of the results that came out of it?
Yes, of course. So from the two studies that I previously mentioned,
we noticed a pattern of sending a lot of links, URL links that leads to different
(34:12):
websites, not necessarily all of them we could identify initially.
So we decided to take all those links, create a list of all of them,
and then run forensic analysis on them.
Forensic analysis, the tool that we use, just take the links and run it through multiple antiviruses.
And based on the classification, if one antivirus classified this link as either
(34:38):
suspicious, malicious.
Phishing, or other, then we, or nothing, if it was nothing, if it was come clean,
that's when we usually took the link and we open it in sandbox. box.
The sandbox is like a secure computer that even if the link is infected,
then it won't impact the entire system, just this particular computer.
(35:02):
And we checked each one of the links and what we found was that 19% were malware,
5% were phishing, and then 5% led to pornographic content, and then And 41% led us to video calls.
(35:25):
And out of the 41% that led us to video call, 85% of them were of a specific
platform called Whereby.
And it was very intriguing because why offenders would use malware,
why they would use phishing attack, kind of was understanding for us.
Like malware can be used because they can infect your computer and then they
(35:49):
can gain access to your camera. They can gain access to your images,
they can gain access to your contacts, your emails, everything, right?
So they can gain access to your computer.
The phishing can provide more information that would then help them manipulate
or coerce the child into more, you know, activities.
(36:11):
Then the video call was, okay, yeah, they use video call, but why specific platform?
So we decided to kind of like look into this platform.
And when I looked at the website, I've noticed that they openly advertise the
fact that they allow embedding code and conversations.
(36:32):
And then when you look at the type of codes that you are allowed to embed,
it specifically states that you can toggle.
Toggle means control a webcam either to open or close, along with opening and
closing the microphone or recording. boarding.
And so I was like, okay, let's test it.
(36:52):
So we actually took the code and we tested it and that worked,
but then we wanted to take it a step further.
So we coded, we added into the script, the possibility that the camera would
open and close every 50 seconds automatically.
And it also worked. So that kind of like made us understand why this platform
(37:14):
would be more killing because once the child is on the.
So what happens is once you log in, it's like any other platforms.
So this platform is web-based. So you don't have to download anything to your computer.
But then once you give it permission to access your webcam and microphone once,
unless you delete your cookies, like everyone always recommends to us,
(37:36):
which miners don't know how to do, it gives the platform complete access to your camera.
And it doesn't have to ask you again if you want to open or close or anything. thing.
It just completely can close and open the camera automatically.
So even if the child doesn't want to, the offender can just open the camera
because they can control it through this platform.
(37:59):
Maybe without even, you know, initially the child do not know what he's up to
and it's without their knowledge.
And if they do not close the computer and they keep the platform open and the offender
can keep opening and closing the camera and recording and taking screenshots
and basically producing child pornography with this, without the child's knowledge.
(38:23):
So I think that this not only shows us that they use certain platforms,
but also they choose the platforms based on their abusability.
That's how I call it. How much they can abuse the platform to abuse the minor.
How How much they can exploit the capabilities and what the platform provides
(38:46):
to them so they can abuse the miner later on is what makes it more appealing for them.
Now that's related to many, to theories from information security about like
why we choose to use certain platforms over others that can explain,
you know, cognitive explanations to this as well.
The easy with which they can do it because the information is openly advertised, right?
(39:10):
So it makes it more appealing, even more to any person who do not possess technical skills to do it.
It doesn't need to have this because it's already there. I was going to say,
it's like its own form of malware from how you were describing it.
That seems like it's probably even easier.
Yeah, but to write a malware, you need to have technical skills.
(39:32):
You need to have coding capabilities. Here, you just have to copy-paste.
And it's less technical than writing a complete malware to embed in a computer.
And that's what we're currently working on. I mean, we published it in The Conversation,
but we're currently publishing the scientific paper behind it.
(39:53):
And we thought it should be alerted because it's such a concerning thing to
know that this kind of platforms exist.
I mean, one of the questions I keep asking myself is for what purpose we design
platforms that allow to open and close camera,
even not just for minors, even us adults sitting in Zoom meetings.
(40:18):
I don't see a need for the host to be able to open and close a camera of a guest in work meeting.
It was very creepy. It always brings me to the question that we need to think
about security by design.
When we design platforms, we need to think about how secure or how much this
(40:40):
platform can be exploited and how we can mitigate those exploitations.
Not just for minors, in general speaking, I mean, any user should be protected.
And that brings questions about who should be responsible for that.
Is it the software companies?
Others? I don't know. It brings a lot of questions that I still don't have the
(41:02):
answers for because I keep thinking about this kind of stuff.
But it questions for policymakers to consider as well. Absolutely. Absolutely.
Because the police can do as much as they can, especially with the limited knowledge
they have on technology.
But then we have companies that possess a lot of knowledge and power,
(41:23):
but then they choose to design this kind of platforms.
It feels like it's crossing a line. And it takes me back to all of those movies,
where you see randomly the little light on someone's laptop or whatever goes
off, and they don't realize it and things are unfolding and then they realize.
And now you have me wanting to go in and delete my cookies immediately because
(41:47):
I don't think I've done that in a while.
But yeah, it does bring up a lot of those questions, which yeah,
I don't know if we'll ever... Well, hopefully we have answers for them at some point.
All right. So given everything that we've discussed, right?
We've covered a lot of ground here all around the same topic,
but we've talked about guardianship, sexual knowledge versus being naive on
(42:09):
the topic, online groomers using web cameras, especially through this website called Whereby.
You've also mentioned the importance of some of your work for things like sting
operations or other elements of the police and court processes.
But what can you give us some future directions for research and then maybe
(42:30):
delve a little bit more into the policy implications of your work.
Yes, of course. So in terms of policy, I think I mentioned a few earlier,
either for training parents.
Training minors, parents, particularly on like the need for them to supervise
their children when they go and spend time online.
(42:51):
And then in terms of sexual knowledge, we need to train children to not discuss
even their sexual knowledge online, not just their experiences or anything, just avoid it.
And we also need to, so some of the policy implications I already mentioned previously.
(43:12):
One of them is the training awareness that we not only need to train minors,
but we also need to train parents about how to actively supervise their children
when they go online and spend time online because that has an impact on the offender's behavior.
Once they know the parent is around, they would stop doing those,
stop the grooming process.
(43:33):
It will deter them. Then there is the training for the children to know that
they shouldn't discuss either their,
Not just their experiences, but also not discuss their knowledge on the topic
in general, because that makes them also at risk for the grooming process and for offline contact.
(43:54):
More importantly, then the implication for police operations and how that can
contribute for sting operations, as I mentioned previously.
It also has theoretical contributions, demonstrating that theories that we use
for traditional crimes can be applied for cybercrime and contribute for our
understanding of cybercrime.
(44:16):
And in terms of future research.
Currently conducting an experiment looking at the prevalence,
but also the differences in which offenders approach different genders,
particularly focusing on female,
male, and non-binary victims to see whether there is more use of coercion,
(44:40):
aggression towards different genders, how they go around. Maybe it's similar.
We don't know. The thing is, we don't really know. There's not much literature around this.
And also understanding the prevalence of online grooming among the LGBTQ plus
community, which is very important because at the moment we have no information about it.
(45:01):
No to very minimal information about it.
So we cannot really fully understand the extent and the severity of the issue among this community.
And then also, I'm trying
to increase the database that I have of grooming conversations to be able to
train an AI model that will allow us to identify grooming conversations sessions
(45:27):
before the unfold and to be able to either block or stop those,
depending on the platforms that in the future would be willing to work with us.
That's kind of where my end goal is to develop a tool that will allow us to
identify the rooming conversation before it even happens.
(45:50):
I feel like that's a big goal, but a really important one. And it sounds very,
very cool. Cool. I hope you have a lot of success at that.
So I think it would make a very big impact. Thank you.
All right. Well, thank you so much, Eden. Those are all the questions that we had for you today.
We really appreciate you taking the time out of your day to talk with us.
(46:10):
If people have any more questions for you regarding your research,
where can they find you? What's the best way to reach you?
Is it email, Twitter, text, something else?
So first of all, thank you very much for having me. I really enjoyed this conversation
and talking about my research and talking about the importance of sharing it.
(46:31):
So thank you guys for doing this podcast as well.
And then if anyone has any questions, they can contact me on my email.
And I have an ex-account as well, which is Eden Kamar.
And then I have LinkedIn that people can find me there as well.
And I'm happy to answer any questions, really.
(46:54):
Awesome. Well, thank you again. Really appreciate it. Yes, thank you.
And it was so great meeting you.
Hey, thanks for listening. Don't forget to leave us a review on Apple Podcasts
or iTunes, or let us know what you think of the episode by leaving us a comment
on our website, thecriminologyacademy.com.
You can also follow us on Twitter, Instagram, and Facebook at The Crim Academy.
(47:15):
That's T-H-E-C-R-I-M-A-C-A-D-E-M-Y. or email us at thecrimacademy at gmail.com. See you next time.
Music.
Music.
(00:14):
Hi everyone, welcome back to the Criminology Academy podcast where we are criminally academic.
My name is Jen Tosley. And my name is José Sánchez.
Today we have episode 93 and for this episode we have on Dr.
Eden Kumar who will speak with us about cybercrime, specifically the online grooming of children.
Dr. Eden Kumar is a postgraduate in the Evidence-Based Cybersecurity Research
(00:39):
Group at Georgia State University.
She received her PhD in Criminology specializing in in Cybercrime and Cybersecurity
at the Hebrew University of Jerusalem and her MPhil in Chronological Research
from the University of Cambridge.
Her research focuses on cybercrime and cybersecurity, namely technology,
facilitated sexual abuse of minors, phishing attacks, illicit markets,
(01:01):
malicious codes, and information system vulnerabilities.
In this episode, our conversation centers around two articles and one op-ed.
The articles, The first one is Parental Guardianship and Online Sexual Grooming
of Teenagers, a Honeypot Experiment Published in Computers and Human Behavior in 2022.
(01:22):
And the second article is The Relevance of Targets, Sexual Knowledge and the
Progression of Online Sexual Grooming Events, Findings from an Online Field
Experiment Published in Justice Quarterly in 2023.
Both articles were co-authored with David Myman, David Weisberg, and Diko Shaba.
(01:42):
We're also going to discuss op-ed in the conversation co-authored with C.
Jordan Howell titled Online Predators Target Children's Webcams Studies Fine, published in 2023.
So with that being said, let's bring Eden in. Hi, Eden.
Thank you so much for joining me and Jose today. We are excited to speak with you about your work.
(02:04):
Thank you for having me today and for inviting me me to be on the podcast.
It's great and a pleasure to be here with you guys. All right. So let's get started.
We're going to go over some background first, kind of get the lay of the land
when it comes to your research area.
And so we have had a couple of episodes in the past talking about cybercrime,
(02:24):
specifically Zoom bombing and the dark web.
And so in this episode, we're going to be speaking about something related but different,
which is online grooming. And so we want to just set the stage and begin with
a definitional question, which is our specialty here on this podcast.
And so what is online grooming? So it's very difficult to define it because
(02:48):
there's many different definitions and common ones that are used often refer to offline grooming,
which refers to grooming as any manipulation.
Manipulative behavior towards children and the child's environment to convince
them to do any sexual behavior.
Now, this changes when the behavior moves to the online environment because
(03:11):
the parents or the environment of the child is kind of different.
They're not around there. They don't know. They don't see. They're not aware
of what's going on usually.
So what What happens is the entire manipulation is toward the child.
And so the definition that I adopted, which is by Joblay et al.
(03:31):
2021, she defines it as any interaction between an adult and a minor with the
intention of sexual interaction, either online or offline.
So it's more about the interaction between the adult and the minor rather than
the environment outside of this interaction that influences sex.
Whether sexual offense occurs or not.
(03:53):
So that's how I adopted it and how I see online grooming.
And I think it's an inclusive one because it takes into consideration the shift
between the offline and the online, which previously has been disregarded.
Yeah. And I imagine that this is a difficult thing to study and get estimates
about, but do we know what the prevalence of online line grooming is?
(04:16):
Like how often does this occur?
So there is estimates based on police reports, which has criminologists we know
as the tip of the iceberg.
And in girls, it's one in three, according to the police records.
In boys, it's I think one in nine, if I quoted correctly.
(04:37):
But we don't really know. And we only look at female and male whilst we know
there are other there are genders out there and we don't know the reports towards those genders.
Research that looks around online grooming up until 2019 said that there was
a decline in online grooming.
But since COVID-19, we see a significant increase in the number of cases.
(05:02):
And we also see differences from before to after and more use and manipulation
of technology to commit sexual offenses against minors?
One in three and one in nine. That seems really high already,
let alone if it's the tip of the iceberg, like we typically know police reports are.
Yeah, that's pretty high up there. So there's like a few things that we know
(05:25):
about crime and victimization.
And one of those is that usually for most crimes, it'll involve someone you
know, rather than like a complete stranger.
When it When it comes to online grooming, does that typically hold or who is
most likely to be doing the grooming and what does the usual grooming victim look like?
(05:47):
So in comparison to offline where the victim and the offender have some relationship,
they know each other, either they've been in contact somehow,
either it's a family friend or a coach, sports coach or a teacher,
or someone related from the environment of the minor online,
(06:08):
it's a complete stranger.
That person can be in Australia, in the UK, India, they can be anywhere and
they can be talking to a minor in the US or any other country around the world.
So the difference is not only that it's a complete stranger,
it's also someone that can be miles and miles away and they do not necessarily
(06:33):
have to leave the comfort of their home to commit the crime.
They can just stay in front of their computer.
And achieve whatever goal they intend to achieve at that point.
In like an anonymous fashion too, right? I'm assuming most of the perpetrators would be anonymous.
(06:55):
Yes, and also it's hard to know how much of the information they provide is
true or not true because they want to disguise their identity at least until
they do some risk assessment of who you are and how much they can trust you. All right.
So when Jose and I were talking about online grooming, my mind went to shows
(07:16):
like Criminal Minds or other popular TV shows that have had some element of
grooming, whether it's in person or online.
And so people who are familiar with these shows may imagine someone looking
to groom a child or a teen goes to a chat room and catfishes someone into an
online relationship by pretending to be someone around the victim's age.
(07:40):
Is this actually how online grooming works or how is contact made between the
offender and the victim?
So chat rooms are still quite popular for this type of behavior,
although there are some use of some other platforms nowadays like Facebook, Instagram.
Dating apps for teenagers are also quite common for groomers,
(08:03):
but chat rooms are quite popular.
And the offenders, some of them are still using deceitful behavior and deceiving
behavior and pretending to be minors.
The majority of them do not feel the necessity to do so.
And they just state that they are old or just say, I'm old or refer to themselves
(08:25):
as daddy, which is a nasty way to describe themselves. myself.
So I think that that's going to set us up for the first paper that we're going
to discuss, which is titled Parental Guardianship and Online Sexual Grooming
of Teenagers, a Honeypot Experiment.
And so our first question about this paper is, what was the motivation behind it?
(08:46):
And so what were the gaps that you're trying to address with this paper?
Many studies looked at guardianship in the context of online grooming.
The majority of them used surveys to collect data.
They referred to where in the house the computer was based, like whether how
much involved the parents were, if there was parental conflict in the house
(09:08):
that the child was disposed to.
They looked at different stuff, but I was more curious about how the way that
children are navigating or like using data.
The online environment, and the context of the household that affect the context of guardianship.
So it takes us a little bit to Rosgood's theory when he talks about unstructured socializing,
(09:34):
when he says that when minors are spending time outside without parental supervision,
they're more likely to conduct,
like to be criminal, to do criminal activity.
But they are also more likely to be victims of crimes because there's no one
there to supervise them.
And so relying on that with the context of guardianship in general,
(09:58):
it was interesting for me to see like how the parental supervision of children
who are going online, because the online environment nowadays is like the playground of the 90s.
Children, instead of going outside to play in the playground.
They go online and they play games and they socialize and they meet people.
(10:19):
So the way that we socialize changed.
So we also need to adapt how we look at theories that applied there back then
to how they can apply nowadays to criminality and victimization.
And so I looked at how we can relate the theory into the cyberspace.
And in this context, parental supervision would be a parent sitting next to
(10:42):
the child when they're spending time online.
And that's what I termed active supervision, when parent is actively supervising
their children's online behavior. behavior.
Now, although Osgood didn't refer to it, but there's another theory that I drew
on from Ryan's that he looked at different levels of guardianship in the context
(11:03):
of neighborhoods, and he referred to active, proactive,
and passive guardianship, where passive for him was when someone was just in the house.
Someone was in the house and that prevented someone to burglar the house.
So I wanted to see that in the context of online grooming, the mere presence
of a parent in the house might affect how the groomer reacts.
(11:25):
And how he changes his behavior towards the entire situation.
So what we did is we looked at those on passive or active parental supervision
compared to no supervision at all, which is the most common nowadays that children
are just playing online,
to see how once the offender encounters a different situation,
(11:48):
how he reacts to the situation, and whether it is effective in terms of guardianship.
Because asking parents on a survey or children on a survey how their parents
behave and then refer to it if it affects,
you know, guardianship is not as effective as if we actually talk to the groomers,
actively engage with them, and we see how they react to different scenarios
(12:12):
that we can actually measure the effect of it on the progression of the grooming process.
And so that's kind of how I came up. And it's really interesting,
these different levels of parental supervision and how that impacts these different
interactions and potential grooming situations.
(12:33):
As I mentioned the title of this paper, you mentioned that this is a honeypot experiment.
Can you tell us what a honeypot is? Yeah, so honeypots is an environment to
which we lure an offender to be able to study his behavior, his reaction to different scenarios.
(12:54):
It's kind of like a mousetrap, right? We pretend to be, in the context of this
study, we pretended to be 13-year-old girls.
We created fake users of 13, 14-year-old girls on a sample of 23 chat rooms.
And then we just waited, really, for them to approach and start communicating with us.
(13:15):
And once they started engaging in communication, to ensure that we are actually
talking to offenders and not talking to minors, which is also a bit problematic,
and I'll talk about it further later.
The first question was ASL, which is age, sex, location.
We wanted to make sure that we're talking to adults and not minors.
But then it's also a limitation.
(13:36):
We started talking in the beginning where some of them are using a deceiving
approach of pretending to be be minors.
So we might exclude those who use this deceiving behavior from this study,
but we had to ensure that we are only talking to adults or to individuals who
identify themselves as adults.
(13:58):
And that's why we asked this question initially.
And after we made sure, based on the treatment that was assigned to the user
that the offender approached to, then the conversation would continue based on that.
I've heard of honeypots before in our other episodes, but yours was a little
(14:18):
bit different where you used a honeypot chatbot, which was something new that
I hadn't heard of before.
Can you tell us more about how you used this honeypot chatbot and what about
collecting the 639 communications you therefore used in this study?
Of course. So a chatbot is basically automation of a conversation flow using
(14:42):
natural language processing.
So basically, we have a script. We try to predict any questions that the groomers would ask us.
And then based on this, we created a script or flow of conversation with a set
of answers that we wanted the machine to reply.
And then each treatment had a different conversation flow.
(15:05):
But since it's the first study ever using chatbot in the context of online grooming,
it was very tricky to predict any potential behavior that the groomer would
or any potential question that the groomer would ask us.
So what we did is the bot was functioning in a way that if it wasn't 100% sure
in the question that was asked, it would email me as a person who is supervising the bot.
(15:32):
I can read the email with the conversation flow.
And then, and it will say like the email would be like, the title would be active
or passive or no guardianship.
So I know exactly to which treatment it belongs.
And then I would have the conversation flow. I can read through,
then I can reply through the email, and then the email would be sent automatically,
(15:54):
like the message that I'm sending an email would be automatically appearing
in the chat room and sent to the offender.
And that's how we collected it. At the end of each data collection,
the bot automatically saves everything to an encrypted platform to ensure that
the data is secure and also no one else can access it.
(16:16):
And we anonymize information that is not supposed to appear, any identifiers, and
So that's how it works. And it has also the ability to flag us.
So when we designed the bot, I wanted to be able to know which conversations
had keywords that I know are groomers' words, usually.
(16:40):
So it will flag this conversation.
So it would just basically drop me an email saying, red flag on a conversation.
So I can go to the chat room or like go to the machine where the bot is running
and read through when bot is still communicating.
The bot was running through. So we had in total, we were sitting on 23 chat rooms.
(17:03):
We had three users on each and we were running it through 14 virtual machines.
So we needed a lot of power and we
had about three emails in total
communicating communicating with me and the
bot communicating with itself so it was a
huge operation to make this all work but
(17:24):
it's a cool way not only to collect data
but also to save time because if
I had to do it without the chatbot I
say it always is a joke but I would retire before I had
a piece because it was so
much time consuming to communicate with each one
of them individually So having the machine answering automatically based on
(17:47):
those scripts and me just supervising whenever and assisting the bot whenever
it's not sure allowed me to collect 639 conversations in about three and a half months,
which without it would take me probably a couple of years.
Yeah, no kidding. That's really cool. I imagine it still took you quite a bit
(18:10):
of time considering this is the first time, like you said, this chatbot was used.
Did you have to intervene like when it was having trouble figuring out what
to say a lot or was it pretty on top of the conversations? conversations?
Unfortunately, yes, particularly in the control group.
So in the no guardianship at all, because the conversation went on and on and on and on.
(18:34):
And it escalated to points where I wasn't expecting it to escalate.
And so I had to intervene because I didn't foresee this going on so long.
So that's mainly when I had to intervene quite a bit.
And that was also challenging for me those conversations
that went on and on because it escalated to
(18:57):
points where it got very nasty but just
put it like I can imagine but I don't want to yeah so just to break this down
and be more specific so you had your two treatment groups then in one control
group was the only thing that differed between these the active passive and
no parental supervision Or were there other things too?
(19:19):
Just this. I mean, they had different... So the way we designed it was that
I went to this website. I don't know if...
Anyone mentioned it before, there's this website that calls this person doesn't exist.
So it's an AI that develop images of, it takes a lot of images from online and
merge them into one image of a person who doesn't exist.
(19:41):
And so I went on this website and I downloaded images of people who appeared to be 13 year old girls.
And I use those images instead of, you know, real people. And then we had random usernames.
So literally just Googled girls' usernames to chat rooms and ideas.
(20:04):
And then I had a list of whatever people previously used.
So I just scrapped and used those just to make it look like a 13-year-old girl's chat room profile.
And then what we manipulated was the answers.
But then that was tricky as well. So initially, I thought that based on all
(20:25):
the literature and what we know about grooming, that offenders are doing risk
assessment and they ask, hey, your parents are around. Are you alone? Are you in your room?
Where are you? Something, right? So out of the 639 conversations,
we found that about, if I remember correctly, about 57 conversations,
only in 57 conversations, they actually asked.
(20:47):
Class and it was shocking because I
thought that's going to be one of the most prominent questions
so in the first few days I had to kind
of change the design because we designed it so okay whenever
they ask we reply but then when we notice they're not asking I was like okay
either the literature something in the literature is not as it was supposed
(21:09):
to be or something is off and then we were like okay so how are we going to
move with it so we decided that whenever Whenever they started asking sexual questions.
Questions, that's when we're going to introduce our treatment.
So when they ask, for example, something that is not too nasty,
(21:29):
they would ask, have you ever kissed?
Or what are you wearing? If it's the active treatment, the bot would trigger, I'm not alone.
My mom is next to me. I'm not allowed to be on the computer on my own.
And if it was the passive one,
it would be, be right back my mom is
calling me from downstairs so they would know the mother
(21:52):
is in the house and then we would wait five minutes and then we message back
back and see like if they
pick up or they drop it and we hypothesize
that the active parental guardianship would have more impact and would be more
likely to reduce the grooming process from continuing and even when we compare
(22:15):
compared to the passive guardianship.
And the results indicated that exactly.
We found that 92% of the conversations, so the likelihood was reduced in 92%
compared to the control group.
And then we had 57% reduction for the passive.
(22:38):
And then there was 87% between the active and the passive. So it was the active
parental supervision was substantially reducing the likelihood that the grooming
process would continue.
But then something that wasn't on the paper, those who did continue the grooming
process shifted the abuse from the minor to both the minor and the parent.
(23:05):
So they weren't just satisfied with abusing the minor. They were like,
okay, so does your mom want to take part of it as well? Wow.
Yeah. Or questions like, is your mom old as you? Okay. Interesting. Interesting.
I wouldn't have expected that. Yeah. So, you know, it goes back to the literature
that talks about the fact that some sex offenders are just sex offenders in
(23:28):
general. It's not just against minors.
It's also against, you know, they perpetrate against adults as well.
They do not differentiate.
They don't have preference necessarily.
So they just combined the two in this case and shifted to both the parent and the minor.
Well, parent. the big parent in this situation.
(23:52):
But it was a bit surprising to me to actually see it.
I think the first few times I was very surprised that's the case.
Do you know how prevalent that was? How often were they like,
bring your mom into it too?
I never quantified it, to be honest.
(24:13):
I don't want to give them numbers because I don't know the exact numbers,
But it wasn't uncommon for those who weren't deterred by the guardianship.
Right. It was enough to where it was kind of a pattern, just like a one-off.
Yeah. All right. So just to quickly recap. So you found that having the active
guardianship really reduced the risk for grooming.
(24:37):
And then that was followed by the passive guardianship. And then last was no
guardianship. It was like in comparison. The first model was in comparison.
So we compared active with no guardianship, passive with no guardianship,
and then we compared the active to passive.
And in all cases, in both two models, the active and the passive was more likely
(25:01):
to reduce the online grooming than the control group.
And then when we compared the two, it was the active that was more effective
in reducing the likelihood of grooming from proceeding.
All right. So you sent us another paper too.
And so this next one that we're going to talk about is titled,
(25:21):
The Relevance of Target Sexual Knowledge in the Progression of Online Sexual
Grooming Events, Findings from an Online Field Experiment. So same question we asked you last time.
What was your motivation or why did you decide to study this and write this paper?
So there is this assumption that naive children are more likely to be victimized.
(25:44):
And I read a few papers that talk about how sexually knowledgeable offenders
that were interviewed says that they actually look for those who are sexually
knowledgeable compared to the naive.
And I decided to test it. I mean, let's, why not?
I mean, we need to make sure that we differentiate between sexual knowledge.
(26:09):
And sexual experience. It's two different things. Sexual knowledge is me knowing
certain terminologies, understanding what it means.
Sexual experience, it means you have tried it, have done it before.
We don't test this. We test only the knowledge.
So whenever there was a situation, if we were asked if we tried something, it will be no.
(26:31):
But whether we know or not know, No, it's a completely different thing.
And I think this is a very relevant thing for us to consider.
We want to develop awareness trainings, right?
We want to prevent and we want to make the children aware of the risks.
And one of the things we need to consider is what we're teaching.
(26:51):
Are we teaching them to avoid to talk about certain things because it's riskier?
Or actually, if they talk about it, it might not have any effect.
So for us to exactly know how to develop trainings, we need to know that.
So that's when I kind of decided to study the effect of sexual knowledge on the grooming process.
(27:14):
And what we did is we used the same approach as in the first experiment.
But this time, we only had two users on a sample of, I think,
21 chat rooms, if I remember correctly.
And we manipulated whether the child knew what certain words mean or they did
not know what it means and see how the groomers react to this particular scenario.
(27:41):
And in this case, we found that actually, compared to what everyone's saying
about naive children being more at risk of grooming,
the sexually knowledgeable children were much more at risk of grooming and were
more likely to receive requests for offline contact,
which is much more alarming than the actual previous findings because of the
(28:06):
physical and psychological impact that offline sexual intercourse can have.
Sexual abuse has, not minimizing the online one, because online one is substantial
as well, particularly with reliving the victimization every time an image is
shared or reposted or whatever.
Both of them are bad equally, but I just think that physical abuse is taking
(28:31):
it a bit step further. And so...
That was very important, I think, finding for us to then be able to not only
think about like awareness trainings, but also think about sting operations,
because a lot of the times sting operations, police sting operations are like
(28:53):
it gets to the point where it gets to court, but then the court dismiss it with
the accusation that the conversation was leading.
Officer was leading the groomer, and therefore the evidence is dismissed.
And so we demonstrate with this study that just by saying that we understand
certain terminologies, sexual terminologies, without even having any sexual
(29:18):
experience, still lead to an offline encounter. counter.
So if sting operator officers just use this kind of approach,
they might be able to prosecute the offender without having evidence dismissed
later on when it gets to court.
So I think it has substantial contribution for both areas.
(29:43):
Yeah, absolutely. I want to go back to this point that you said about this request
for an an offline meeting and a point that you made earlier where online grooming
the offender and the victim might be,
you know, miles, countries apart.
Did that play any role into this request for an offline meeting?
(30:05):
Or maybe you haven't looked at this at all. And I'm just hitting you with a random question.
But I think it's interesting, this offline meeting request, when potentially
one person could be in Australia and the other one in Canada.
So I haven't looked at it to answer to your questions.
I haven't looked at it. To one of you that I do remember,
(30:29):
I remember the chatbot was in New York and the offender was saying that he was
in the East Coast as well in one of the conversations that I do remember.
But I haven't looked at it in general to provide more.
But it's not uncommon for online groomers to travel.
It is quite common. And for example, in the UK, one of the ways that the National
(30:55):
Crime Agency managed to prosecute online groomers is when they land in the UK.
So they communicate with them to the point where they arrange offline meeting.
And once they land in the UK territory, in the airport, they arrest them in the airport.
So it's not uncommon for offender to be in trouble.
(31:17):
Interesting. So, one of the things that you cover in the paper is like the process
of grooming or like these various phases of online grooming that groomers tend to go through.
How were you trying to identify these phases or like flagging like,
okay, we've hit like a...
(31:38):
I should have written down the names. Like the report building,
sexual content assessment, those phases, were you keeping an eye out for trying
to track where you were? So there's different types of like grooming process, right?
Typologies there is like there
isn't really one that is agreed on
(31:58):
i mean there is a connell's one that it's the
most cited one that talks about report building and
then moving on from one case to another in a
linear way but from my studies i've noticed
that it's not as linear as people think it is it's not like the groomer start
a conversation with building report they might just go straight into what they're
(32:20):
after and like send you a link to a video call and ask you to jump on the call
and let's just do whatever day he wants to do on the video call.
Or he immediately start with saying sexual things without building any initial report.
But then when, for example, we always had to ask, no matter what is the first
(32:44):
question that the groomer asked or the first message that the groomer asked,
we had to kind of like take it a step back to ask about the age.
So it would be like, okay, now after I already asked, she asked me about the age.
So now I need to, maybe I should start building a bit of rapport with her before I move on.
Or they might just answer the question and then immediately send the link again,
(33:06):
like just come to the video, like to the video call, just come to the video call.
So it's not linear or progressing in a way that we can describe.
What we do find is that something that is not really being discussed in previous
literature is that there are other pathways into grooming that do not appear in the literature,
(33:28):
like the use of URLs that lead us to websites that can be utilized to different things.
Okay, so I think that can get us into the last piece that you sent us that we're going to discuss today.
So in May of 2023, you wrote a piece for The Conversation about online groomers and using webcams.
(33:50):
And the data that you presented in this article is part of a study that's currently under review.
Can you give us an overview of this study and sort of the results that came out of it?
Yes, of course. So from the two studies that I previously mentioned,
we noticed a pattern of sending a lot of links, URL links that leads to different
(34:12):
websites, not necessarily all of them we could identify initially.
So we decided to take all those links, create a list of all of them,
and then run forensic analysis on them.
Forensic analysis, the tool that we use, just take the links and run it through multiple antiviruses.
And based on the classification, if one antivirus classified this link as either
(34:38):
suspicious, malicious.
Phishing, or other, then we, or nothing, if it was nothing, if it was come clean,
that's when we usually took the link and we open it in sandbox. box.
The sandbox is like a secure computer that even if the link is infected,
then it won't impact the entire system, just this particular computer.
(35:02):
And we checked each one of the links and what we found was that 19% were malware,
5% were phishing, and then 5% led to pornographic content, and then And 41% led us to video calls.
(35:25):
And out of the 41% that led us to video call, 85% of them were of a specific
platform called Whereby.
And it was very intriguing because why offenders would use malware,
why they would use phishing attack, kind of was understanding for us.
Like malware can be used because they can infect your computer and then they
(35:49):
can gain access to your camera. They can gain access to your images,
they can gain access to your contacts, your emails, everything, right?
So they can gain access to your computer.
The phishing can provide more information that would then help them manipulate
or coerce the child into more, you know, activities.
(36:11):
Then the video call was, okay, yeah, they use video call, but why specific platform?
So we decided to kind of like look into this platform.
And when I looked at the website, I've noticed that they openly advertise the
fact that they allow embedding code and conversations.
(36:32):
And then when you look at the type of codes that you are allowed to embed,
it specifically states that you can toggle.
Toggle means control a webcam either to open or close, along with opening and
closing the microphone or recording. boarding.
And so I was like, okay, let's test it.
(36:52):
So we actually took the code and we tested it and that worked,
but then we wanted to take it a step further.
So we coded, we added into the script, the possibility that the camera would
open and close every 50 seconds automatically.
And it also worked. So that kind of like made us understand why this platform
(37:14):
would be more killing because once the child is on the.
So what happens is once you log in, it's like any other platforms.
So this platform is web-based. So you don't have to download anything to your computer.
But then once you give it permission to access your webcam and microphone once,
unless you delete your cookies, like everyone always recommends to us,
(37:36):
which miners don't know how to do, it gives the platform complete access to your camera.
And it doesn't have to ask you again if you want to open or close or anything. thing.
It just completely can close and open the camera automatically.
So even if the child doesn't want to, the offender can just open the camera
because they can control it through this platform.
(37:59):
Maybe without even, you know, initially the child do not know what he's up to
and it's without their knowledge.
And if they do not close the computer and they keep the platform open and the offender
can keep opening and closing the camera and recording and taking screenshots
and basically producing child pornography with this, without the child's knowledge.
(38:23):
So I think that this not only shows us that they use certain platforms,
but also they choose the platforms based on their abusability.
That's how I call it. How much they can abuse the platform to abuse the minor.
How How much they can exploit the capabilities and what the platform provides
(38:46):
to them so they can abuse the miner later on is what makes it more appealing for them.
Now that's related to many, to theories from information security about like
why we choose to use certain platforms over others that can explain,
you know, cognitive explanations to this as well.
The easy with which they can do it because the information is openly advertised, right?
(39:10):
So it makes it more appealing, even more to any person who do not possess technical skills to do it.
It doesn't need to have this because it's already there. I was going to say,
it's like its own form of malware from how you were describing it.
That seems like it's probably even easier.
Yeah, but to write a malware, you need to have technical skills.
(39:32):
You need to have coding capabilities. Here, you just have to copy-paste.
And it's less technical than writing a complete malware to embed in a computer.
And that's what we're currently working on. I mean, we published it in The Conversation,
but we're currently publishing the scientific paper behind it.
(39:53):
And we thought it should be alerted because it's such a concerning thing to
know that this kind of platforms exist.
I mean, one of the questions I keep asking myself is for what purpose we design
platforms that allow to open and close camera,
even not just for minors, even us adults sitting in Zoom meetings.
(40:18):
I don't see a need for the host to be able to open and close a camera of a guest in work meeting.
It was very creepy. It always brings me to the question that we need to think
about security by design.
When we design platforms, we need to think about how secure or how much this
(40:40):
platform can be exploited and how we can mitigate those exploitations.
Not just for minors, in general speaking, I mean, any user should be protected.
And that brings questions about who should be responsible for that.
Is it the software companies?
Others? I don't know. It brings a lot of questions that I still don't have the
(41:02):
answers for because I keep thinking about this kind of stuff.
But it questions for policymakers to consider as well. Absolutely. Absolutely.
Because the police can do as much as they can, especially with the limited knowledge
they have on technology.
But then we have companies that possess a lot of knowledge and power,
(41:23):
but then they choose to design this kind of platforms.
It feels like it's crossing a line. And it takes me back to all of those movies,
where you see randomly the little light on someone's laptop or whatever goes
off, and they don't realize it and things are unfolding and then they realize.
And now you have me wanting to go in and delete my cookies immediately because
(41:47):
I don't think I've done that in a while.
But yeah, it does bring up a lot of those questions, which yeah,
I don't know if we'll ever... Well, hopefully we have answers for them at some point.
All right. So given everything that we've discussed, right?
We've covered a lot of ground here all around the same topic,
but we've talked about guardianship, sexual knowledge versus being naive on
(42:09):
the topic, online groomers using web cameras, especially through this website called Whereby.
You've also mentioned the importance of some of your work for things like sting
operations or other elements of the police and court processes.
But what can you give us some future directions for research and then maybe
(42:30):
delve a little bit more into the policy implications of your work.
Yes, of course. So in terms of policy, I think I mentioned a few earlier,
either for training parents.
Training minors, parents, particularly on like the need for them to supervise
their children when they go and spend time online.
(42:51):
And then in terms of sexual knowledge, we need to train children to not discuss
even their sexual knowledge online, not just their experiences or anything, just avoid it.
And we also need to, so some of the policy implications I already mentioned previously.
(43:12):
One of them is the training awareness that we not only need to train minors,
but we also need to train parents about how to actively supervise their children
when they go online and spend time online because that has an impact on the offender's behavior.
Once they know the parent is around, they would stop doing those,
stop the grooming process.
(43:33):
It will deter them. Then there is the training for the children to know that
they shouldn't discuss either their,
Not just their experiences, but also not discuss their knowledge on the topic
in general, because that makes them also at risk for the grooming process and for offline contact.
(43:54):
More importantly, then the implication for police operations and how that can
contribute for sting operations, as I mentioned previously.
It also has theoretical contributions, demonstrating that theories that we use
for traditional crimes can be applied for cybercrime and contribute for our
understanding of cybercrime.
(44:16):
And in terms of future research.
Currently conducting an experiment looking at the prevalence,
but also the differences in which offenders approach different genders,
particularly focusing on female,
male, and non-binary victims to see whether there is more use of coercion,
(44:40):
aggression towards different genders, how they go around. Maybe it's similar.
We don't know. The thing is, we don't really know. There's not much literature around this.
And also understanding the prevalence of online grooming among the LGBTQ plus
community, which is very important because at the moment we have no information about it.
(45:01):
No to very minimal information about it.
So we cannot really fully understand the extent and the severity of the issue among this community.
And then also, I'm trying
to increase the database that I have of grooming conversations to be able to
train an AI model that will allow us to identify grooming conversations sessions
(45:27):
before the unfold and to be able to either block or stop those,
depending on the platforms that in the future would be willing to work with us.
That's kind of where my end goal is to develop a tool that will allow us to
identify the rooming conversation before it even happens.
(45:50):
I feel like that's a big goal, but a really important one. And it sounds very,
very cool. Cool. I hope you have a lot of success at that.
So I think it would make a very big impact. Thank you.
All right. Well, thank you so much, Eden. Those are all the questions that we had for you today.
We really appreciate you taking the time out of your day to talk with us.
(46:10):
If people have any more questions for you regarding your research,
where can they find you? What's the best way to reach you?
Is it email, Twitter, text, something else?
So first of all, thank you very much for having me. I really enjoyed this conversation
and talking about my research and talking about the importance of sharing it.
(46:31):
So thank you guys for doing this podcast as well.
And then if anyone has any questions, they can contact me on my email.
And I have an ex-account as well, which is Eden Kamar.
And then I have LinkedIn that people can find me there as well.
And I'm happy to answer any questions, really.
(46:54):
Awesome. Well, thank you again. Really appreciate it. Yes, thank you.
And it was so great meeting you.
Hey, thanks for listening. Don't forget to leave us a review on Apple Podcasts
or iTunes, or let us know what you think of the episode by leaving us a comment
on our website, thecriminologyacademy.com.
You can also follow us on Twitter, Instagram, and Facebook at The Crim Academy.
(47:15):
That's T-H-E-C-R-I-M-A-C-A-D-E-M-Y. or email us at thecrimacademy at gmail.com. See you next time.
Music.
Popular Podcasts
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
The Nikki Glaser Podcast
Every week comedian and infamous roaster Nikki Glaser provides a fun, fast-paced, and brutally honest look into current pop-culture and her own personal life.