Q&A: Reality Defender’s CEO on combating deepfakes in financial services

The rapid advancement of generative AI has led to a significant surge in sophisticated deepfake attacks, posing a critical threat to financial institutions. Bobsguide sat down with Ben Colman, CEO of Reality Defender, who provided valuable insights into the evolving deepfake landscape and the strategies being developed to combat it.

Colman, with a background in cybersecurity from his time at Goldman Sachs and experience in national defense, discussed the growing challenges faced by financial institutions and the technological solutions emerging to protect them and their customers.

The deepfake threat landscape

Colman’s experience at Goldman Sachs provided him with a firsthand view of the increasing sophistication of cyberattacks.

“AI is really taking over everything, and that includes cybersecurity,” says Colman, noting the shift towards programmatic hacking attempts that leverage AI to impersonate individuals and facilitate fraud on a large scale.

Initially, the focus was on detecting AI avatars and virtual humans, but Colman and his team recognized the broader potential for AI-generated fraud. “We were right,” he says, “We were just a little bit early.”

Currently, financial institutions are primarily contending with AI-generated audio in deepfake phone calls, which are used to impersonate individuals, bank employees, or company leadership to authorize fraudulent transactions or gain unauthorized access to accounts.

Colman warns that the threat landscape is expected to evolve, with AI-generated video becoming a more significant concern as the cost of computing decreases.

The increasing accessibility of AI tools has dramatically altered the threat landscape. Colman illustrates this by pointing out how easily even young children can use readily available apps to generate fake voices or manipulate faces.

“In the wrong hands, this can create risks unlimited by your imagination,” he cautions, highlighting the rise in non-consensual AI-generated content as a serious issue.

Real-time detection implementation

Financial institutions are increasingly implementing real-time deepfake detection to enhance their existing security infrastructure. Traditional security measures that rely on matching personal data are proving inadequate against AI-enabled attacks.

Colman emphasizes that , “the fact that you’re confirming your identity just means that somebody is confirming they have your identity.”

Solutions like Reality Defender offer an additional layer of security by analyzing audio and video streams to detect AI generation or manipulation.

Colman notes that it’s important to view these technologies as part of a broader security strategy: “We are not a silver bullet, but…it’s the most important thing because when all other provenance based approaches are wrong…is right.”

Deploying new technologies within large, regulated institutions presents inherent challenges.

However, the urgency of the deepfake threat is driving faster adoption of these new solutions. According to Colman, companies are actively seeking out these solutions to address immediate needs.

Solutions

Effective deepfake detection often involves analyzing the media itself, rather than relying solely on verifying personal data, which can be compromised. This approach, sometimes referred to as inference, focuses on determining the probability of AI generation. It also detects manipulation by examining various signals within the audio and video.

Companies like Reality Defender utilize multiple detection models and advanced techniques to identify potential threats. Collaboration with generative AI companies also helps these detection providers stay ahead of emerging techniques.

Balancing detection accuracy, performance, and scalability is crucial for solutions in this space. Companies are focused on providing effective and scalable technologies to protect financial institutions and their customers.

Looking forward

The deepfake threat is expected to continue evolving, demanding ongoing adaptation and innovation in cybersecurity. Cybersecurity providers must remain vigilant and proactively update their technology to address emerging threats.

The vision for the future of deepfake detection includes broader protection for individual consumers.

Colman discusses the potential for integrating deepfake detection capabilities into everyday devices and platforms, similar to antivirus software.

“We’re directly working with global device manufacturers, global cloud providers to envision a future where…is just an extension of antivirus,” he states.

Key takeaways

Deepfake fraud presents a widespread risk, requiring a concerted effort to combat it. Given your focus on cybersecurity and security in fintech, it’s crucial to recognize that while the challenges are significant, the development of proactive solutions offers a positive outlook for protecting users from deepfakes.

“Products…can protect users from deepfakes, and we’re excited to expand our solution globally,” Colman concludes.

• Read more under:
• Artificial Intelligence
• Cybersecurity