Meet Ludivine — A romance scammer from Paris
With millions of daily active users and 2B$ in revenues last year, online dating has become the way for people to meet and start relationships. In April 2019, the industry-leader Tinder surpassed Netflix as the top grossing non-game app in the App Store. A strong message to the world: Online Dating is here to stay and thrive.
Today, over 40% of singles use online dating services, and 1 in 5 marriages started online. By 2020 the number will have grown to 1 in 4, and an overall of 310 Mio. people will have turned to the internet to find love.
Online dating sites are an attractive platform, not only for singles trying to meet a life-partner, but also for fraudsters who see easy victims in people looking for love. Unfortunately, as in every success story, the rise of online dating is accompanied by new types of threats to everyone who uses it.
Ludivine from Paris: A fraudster’s fake profile
The term “Catfishing” describes the phenomenon of internet predators that fabricate online identities and entire social circles to trick people. Fraudsters on online dating platforms use different approaches, which range from simple advertisement of illicit services or goods (i.e. spam, prostitution) to advanced schemes, in which a fraudster interacts extensively with his victims to build an emotional relationship and eventually asks for money under a set of excuses. To understand what it takes for online dating companies to stop fraudster and to increase the safety of their users, let us look at the journey of Ludivine. We found this fake profile on a client’s website and stopped it before it could do any harm.
1) Profile creation and Sign-Up
First and foremost, a fraudster needs to create a profile to connect to potential victims. The fraudster hides his true identity behind a deceptive facade, like the 18 years old Ludivine from Paris. Fraudsters steal images on the web to begin the profile setup process. Their images might originate from a foreign Instagram account, escort service, or, as in the case of Ludivine, from a Twitter Account. Today, around 14% of all new profiles on online dating platforms are fraudulent, with little sign of slowing down.
2) Grabbing attention
Ludivine’s pictures and description attracts attention, making her profile appealing to many users. Once initial interest is indicated, many fraudulent profiles quickly (I) start to spam the user with a wide variety of advertisement. Progressive scammers, however, take a more (II) subtle approach. They encourage discussions about personal issues and problems, send seductive images and messages, and try to deeply relate with their victims to exploit them when they are most vulnerable.
3) Moving the conversation off-site
Either (I) earlier or (II) later, Ludivine attempts to obtain a more private method of communication such as an email or phone number. The fraudster will continue building a relationship until the victim is deemed ready for monetary requests like a plane ticket to visit, medical emergencies or worse, blackmailing.
More than 40% of online daters have reportedly encountered a fraudster, not considering the high number of incidents that remain unreported or unnoticed. The victims experience significant emotional distress as well as financial losses. The median-individual-loss to a romance scam reported in 2018 was $2,600, about seven times higher than the median loss across all other fraud types. And the problem is getting worse: Scam incidents have doubled and reported losses increased more than fourfold from 2015 to 2018.
Deep-Dive: The problem of detecting romance-scam
On their search for love, online daters are generally in a critical emotional and sensitive state, which makes them vulnerable. Online dating platforms are typically designed to connect profiles who have no prior contact, and who then start engaging in private chats. Fraudsters exploit this. Today’s widespread access to resources like malware and privacy tools have made it easy to trick most profile verification systems. Platform operators use two content moderation strategies to confront these problems:
- Behavioral-Analysis: Accounts that are controlled by scammers are likely to show differences in their behavior compared to legitimate users. For example, the number of conversations initiated simultaneously, the time waited between the creation of the profile and the first message, and the fraction of the received messages to which the account replies. Such behavior based moderation systems typically identifies words and patterns that are indicative of malicious content.
- Metadata-Analysis: More technically, metadata analysis focuses on drawing conclusions about the trustworthiness of profiles and content from its underlying metadata. For example, the IP-address, geolocation tests, descriptions, keywords, etc. are checked against blacklists to uncover fraudsters.
Lastly, users of online dating platforms are urged to be aware of suspicious signs. If peculiar behaviors are reported and flagged by the users themselves, this can be a powerful source of self-moderation for the online community. Guides provide comprehensive information for online daters about how to recognize, avoid and report scams.
The fraudster’s bottleneck: Image reverse analysis
Even the most progressive and advanced scammer faces one distinct bottleneck on his journey. Every fraudster needs images to create a fake profile.
Finding stolen images = Finding fake accounts 🕵️♀️
Ludivine’s image, for example, was found on suspicious websites with a different associated name and location. The fraudster simply stole images and used them for his fake profile. Commonly, images of attractive young women, such as Ludivine, or handsome middle-aged men, are utilized for a multitude of fake accounts.
Understanding the digital footprints of images across the web is key to uncover fraudsters. Finding stolen images = Finding fake profiles. Combining behavioral analysis and metadata analysis with a thorough reverse analysis of images is the most effective way for content moderation teams to track down fraud and unwanted content.
Powerful and accurate Image-Reverse-Analysis is an unfair advantage for content moderators in the online dating industry. By deriving relevant insights from images, they can boost their content moderation systems, increase user trust and retention, while decreasing manual work.
Navee, a Paris based computer vision company, has built the most accurate Image-Reverse-Analysis tool for the online dating industry. It enables Content Moderators to track down every digital footprint of an image across the web and to automatically label its traits (nudity, emotions, quality). Test the demo here.
Finding and fighting fraud and scams on an online dating platform is best addressed with a combined methodological approach of behavioral analysis, metadata analysis, and Image-Reverse-Analysis. These tools together, enable content moderators to track down threats effectively and to increase the safety of online communities.
You want to protect your online community and verify profiles? Your users have been victim to fraudulent behavior? You need to moderate content at scale?
Let’s start the conversation! We’ll discuss your needs and address them accordingly.