Many of today’s fraudsters are figuring out how to use AI to automate and structure scams that are unique to each person they target. If fraudsters can analyze your data, learn your patterns, track your interests, and exploit who you trust – how can you combat them?
According to SAS Innovate speakers Mauro Souza of SAS and Lucas Palin of GMWIT consulting, the best way to fight AI-powered fraud is with AI itself. “AI for good needs to combat AI for bad,” said Souza, a Systems Engineer at SAS.
AI-powered phishing scams can draw on data analysis, personalization, content creation, scale and automation, explains Souza. “While traditional phishing attacks rely on social engineering techniques, AI-driven phishing attacks rely on machine learning techniques.”
AI-powered phishing scams include:
- Convincing audio and conversational frauds.
- Realistic deepfake content.
- Custom business email compromised attacks.
- Automated posts and profiles on social media.
“We have the fraudsters using AI against us,” said Souza. “We need to do something to fight back.”
Souza and Palin went on to explain how SAS and GMWIT are using generative AI and AI agents to identify and prevent these types of fraud. GenAI techniques for fighting fraud include:
- Identifying rare deviation patterns.
- Simulating multi-vector attacks.
- Proactively analyzing fraud networks.
- Generating synthetic data for training and validation.
- Predicting the impact of potential fraud.
Fighting AI fraud with AI agents
Palin, the Chief Technology Officer at GMWIT, shared a demo of the AI agent his team built to combat fraud using SAS Intelligent Decisioning. The agent was trained using 100,000 examples of phishing emails compared to real emails, he explained.
Plain showed the audience how he asks the agent to analyze new emails, identify phishing scams and explain its findings. Factors in the analysis, according to the agent, include:
- Offers that are too good to be true.
- Grammatical errors.
- Risky domains.
- Urgent tone or request to do something immediately.
Palin said it’s common in Brazil to receive many emails from the government requiring citizens to sign up and register for different things, and fraudsters have copied these emails for phishing scams.
“I can ask AI to analyze it row by row and give me an analysis, said Palin. “It shows me patterns, risks and recommendations.”
Palin demonstrates how easy it is to create a rule in SAS Intelligent Decisioning. “When I create an alert, it’s possible to connect to a database and access SAS Visual Investigator, see all the details and take some actions,” he said.
Even with the best technology, Souza told the audience that education and training are essential to continue fighting fraud. If fraudsters can penetrate your organization through advanced phishing scams targeted at employees or customers, it’s important to train those individuals to recognize fraud and how to use tools to fight it.
