Top takeaways from TrustCon 2023 – Source: securityboulevard.com

TrustCon is THE place to be for trust and safety practitioners to network and learn from others in the industry. While trust and safety covers a broad range of abuses that are seemingly different, such as fraud, child safety, and counterterrorism, many of the strategies to combat abuse at scale are interchangeable, and there were valuable lessons around successes and failures that I learned from fellow peers. 

A combined approach using machine learning models, automated rules, and human review are currently the industry standard, but it’s no secret that generative AI presents some interesting challenges to the traditional way of doing things. Our recent research at Sift shows that in the six months since ChatGPT was released in November 2022, nearly half of consumers admitted to finding it more challenging to identify scams. Interestingly, generative AI is also showing to be effective in aiding trust and safety initiatives as well. 

I also had the opportunity to present my undercover research on Pig Butchering, an increasingly common romance scam involving cryptocurrency. I’ve shared more details on my presentation, as well as overall insights from TrustCon below. 

Danger of dating apps: breeding grounds for fraud

With 50 million Americans now using dating apps to find love and partnership, fraudsters have found new opportunities to spread scams and malicious content across these popular networking platforms. 

One increasingly common romance scam is Pig Butchering, which primarily targets individuals on dating apps, using cryptocurrency as a lure to drain money from victims. Additionally, scammers are known to attack on other types of platforms that allow messages to be exchanged. 

The profiles that target victims all have eerily similar qualities to them—modelesque profile pictures, generic job titles, irrelevant answers to questions, and a tendency to talk about money. 

These scams are predicated on romance, so the first step involves using flattery or the promise of future trips and gifts to draw the attention of their victims. However, this quickly moves on to manipulation, making excuses for why in-person meetings are impossible or becoming insulting and abusive.

Pig Butchering relies on the popularity of cryptocurrency and “get rich quick” schemes. After moving the conversation off-platform, scammers direct their targets to a fake crypto exchange.  The exchange appears to be legitimate, which convinces victims that their investments are real. Once enough money has been drained from victims, these scammers disappear with their profits.

Generative AI: a double-edged sword

TrustCon’s keynote address on opening day focused on the risks and threats of generative AI.

While its accessibility, variability, ease of use, and scale make it the “next big thing” for technology, these are also the very reasons why it complicates trust and safety initiatives. 

At Sift, we’re seeing how fraud is becoming democratized via the deep and dark web, meaning that the everyday individual now has access to tools to commit fraud, without having to be a technically skilled hacker. This inevitably leads to a high volume of attacks, which can quickly outpace manual fraud review.

While we’re well aware of the threats that language learning models like ChatGPT pose due to their ability to generate a high volume of harmful text that’s grammatically correct, and with high variety, it was fascinating to hear how they can also be applied to content moderation. Specifically, language learning models have demonstrated effectiveness when it comes to low-complexity content moderation and policy writing. This has the potential to revolutionize the way teams approach trust and safety, with the obvious advantage of reducing exposure to malicious content on the frontlines. However, language learning models do not perform as well when it comes to understanding nuances that require subject matter expertise.

Ultimately, the consensus is that the best way to fight AI-enabled fraud is with a combination of AI, automated rules, and human moderation.

Fighting fraud with machine learning

Machine learning can analyze large volumes of data in real-time to swiftly identify patterns, anomalies, and trends that are indicative of fraud. For businesses facing these challenges, it comes down to having a network of fraud signals to accurately keep up with evolving fraud. The scale and sophistication of attackers has grown exponentially with the emergence of new technologies, and businesses must set up proactive prevention measures before they quickly become overwhelmed.

If you’re interested in hearing more insights from the Sift team, check out our upcoming industry events and live sessions. 

The post Top takeaways from TrustCon 2023 appeared first on Sift Blog.

*** This is a Security Bloggers Network syndicated blog from Sift Blog authored by Jane Lee. Read the original post at: https://blog.sift.com/top-takeaways-from-trustcon-2023/?utm_source=rss&utm_medium=rss&utm_campaign=top-takeaways-from-trustcon-2023