Incognia uses location intelligence to detect payment scams before money moves

Incognia has launched a fraud prevention system that analyzes historical location patterns and physical relationships between payment senders and recipients to identify authorized push payment (APP) scams before transactions are completed.
The company’s Real-Time Scam Prevention product combines its Physical Social Graph technology with device intelligence, behavioral signals and mule-account detection to assess transactions in real time. Incognia says the approach addresses a gap in conventional fraud prevention, where the person initiating a fraudulent payment is often the legitimate account holder.
“Banks have gotten very good at determining whether the person making a payment is legitimate, but with a scam, the legitimate customer is often the one making the payment,” said André Ferraz, Incognia’s co-founder and CEO. “The question becomes – who are they sending money to, and does that relationship make sense? That is the context Incognia is bringing into scam detection.”
What the solution targets
In an interview with Biometric Update, Ferraz described APP fraud as one of the hardest problems in the industry. “This type of fraud is particularly challenging because it’s the legitimate account holder logging into their account and going through the entire process of sending money to someone who is a scammer,” he said. He cited romance scams on social media, impersonation scams in which someone calls claiming to work for the victim’s bank, and scams in which fraudsters impersonate family members on social media to ask for money. According to the FTC, as cited in Incognia’s announcement, consumers reported losing $3.5 billion to impersonation scams in 2025, nearly triple the amount reported in 2020.
The patented technology behind it
The solution is powered by Incognia’s patented Trusted Locations technology. Ferraz said that it “focuses on understanding users’ physical behavior and using that information to establish an identity for each individual.” The technology “establishes persistent location patterns based on where users regularly spend time.”
Ferraz said the company’s proprietary location technology can pinpoint a device’s position with an accuracy of less than nine feet. “It’s multiple times more precise than GPS,” he said. “Even within an apartment building, for example, we’re able to distinguish between different floors and different apartments.”
Incognia built what it calls the Physical Social Graph on top of that technology. “We’re essentially building the equivalent of a social network that forms automatically based on the physical relationships between users,” Ferraz explained. He described phones in the same location detecting the same Wi-Fi signals, Bluetooth devices and other environmental signals, which indicates an in-person interaction. The graph builds connections based on physical proximity and then determines the strength of each connection. “The people I live with, my family, have devices that are in the same location multiple times throughout the week. We understand that this represents a very strong connection,” he said. His relationship with coworkers, whom he sees in the office a few days a week, is “not as strong as the relationship I have with my family.” The graph identifies first-, second- and third-degree connections and beyond. Rather than relying on a snapshot of where a customer is when a transaction occurs, the solution compares historical location patterns between the sender and recipient.
Alan Gomes, Incognia’s CTO, said: “Two people can live in the same building and have no real-world relationship, while two people living in completely different cities may know each other well. That’s why distance alone isn’t enough. Our apartment-level indoor location intelligence identifies patterns and connections over time, giving banks context that IP addresses or GPS coordinates alone cannot provide.”
Why banks have struggled to stop scams
When the bank analyzes an APP fraud transaction, Ferraz explained, everything looks fine. “It’s the right person, using a trusted device, potentially from a trusted location,” he said. “All the signals the bank has on the sender’s side check out, so it proceeds with the payment.”
“For a long time, the industry almost reached the point of giving up on this problem,” Ferraz said. He added that even fraud prevention vendors started placing the blame on social media companies, telecommunications providers and others that, in his view, “don’t have much to do with the financial services ecosystem.” Regulators have stepped in. In the UK, banks are required to reimburse scam victims, with the sending bank usually paying 50 percent and the receiving bank paying the other 50 percent.
Most banks use a combination of signals, including device intelligence, behavioral biometrics, facial biometrics and liveness detection. “But in most cases, they’re only verifying the person sending the money,” Ferraz said. “They don’t have much intelligence to analyze both sides of the transaction.” Even when they do, he said, they lack effective ways to establish the relationship between sender and recipient, and can generally only check whether the recipient has previously been flagged for fraud or risk. “But fraudsters move very quickly,” he said. “They’re constantly creating new accounts.”
What the data shows
According to Ferraz, the company analyzed 230 million transactions over one month within three banks, where it could see both sides of the transactions, using labels from the banks identifying each transaction as legitimate or fraudulent. Eighty percent of legitimate payments were between users with a first-degree connection. “In other words, they know each other in the physical world,” he said. Fewer than 3 percent of payments were between users who did not have even a third-degree connection. “We found that less than 3% of payments fell into this category, but 65% of scams were concentrated in this group,” Ferraz said.
The Physical Social Graph technology alone identified approximately 70 percent of scam transactions in the analyzed dataset, demonstrating “how historical real-world relationships can provide valuable evidence of scam risk, even when conventional fraud signals appear legitimate.”
“Incognia uses approximately 500 other signals to detect fraud and scams, but this is by far the most relevant,” Ferraz said. “We believe we’ve finally cracked the code for scams.”
How it works: a Zelle payment
Ferraz walked through a Zelle payment to a friend. The customer opens the banking app, goes to the payment menu and enters the recipient’s information, such as a phone number or account number. “At that stage, the bank calls Incognia’s API and asks: What’s the risk associated with this transaction?” he said.
Incognia’s AI model combines Physical Social Graph intelligence with mule-account detection, active-call detection, remote-access-tool detection, device integrity and location-spoofing detection to assess payment risk in real time. On the sender’s side, Ferraz said, Incognia analyzes whether the device is actively on a phone call, whether it is infected with malware and whether remote access tools are running. “But more importantly, we analyze the recipient,” he said.
To determine whether a recipient may be a mule account, Incognia looks at whether the device currently operating the account differs from the device used to create it and behaves completely differently, which “could indicate that different people are operating the same account.” It also looks at whether the device is located somewhere associated with a fraud ring. “We recently identified a case at a bank where 150 devices operating from a single apartment were connected to 50,000 bank accounts at the same financial institution,” Ferraz said.
Ferraz gave an example. If Incognia identifies that the recipient is potentially operating a mule account, and the sender’s relationship with that person is only a fourth-degree connection, “we would tell the bank that the risk of this being a scam is extremely high and that it should take action.” The bank could then block the transaction, delay it, request additional verification, call the customer or approve the payment. “The bank owns that decision,” he said. “Our role is to provide the intelligence indicating the likelihood that the transaction is a scam.” Banks use these assessments alongside their own data and policies to determine the appropriate action.
First-party fraud: a common thread with Socure
Biometric Update recently covered a Socure study, and some of its findings overlap with Incognia’s, particularly around first-party fraud. First-party fraud is becoming increasingly important across fraud categories, and traditional identity verification signals, whether document verification, facial biometrics or other forms of authentication, show that the person conducting the transaction is legitimate.
“What’s particularly interesting is that the same solution we found for APP fraud also helps address first-party fraud,” Ferraz said, calling it the second finding in the company’s report. Approximately 10 percent of reported scams involved users living in the same household. “That’s clearly indicative of first-party fraud,” he said. “The sender and recipient live in the same apartment.” Ferraz said these individuals transfer money to someone living in the same household, then call the bank, claim they were victims of a scam and get reimbursed.
Asked how banks are solving this problem today, Ferraz said: “They’re not.” He considers APP fraud and first-party fraud “two of the major unresolved problems in our industry. But there’s now a solution for that.”
Industry collaboration and privacy
Fraudsters collaborate and share loopholes across their networks. Asked whether the industry is doing the same, Ferraz said: “I’d say fraud prevention vendors aren’t collaborating very much. But banks are willing to collaborate through fraud prevention vendors to share fraud intelligence.”
The most important issue, he said, is that banks don’t want to share detailed personal information about their customers with other banks. Incognia says it anonymizes the data so it can build the network without ingesting users’ raw personal information. “We don’t need to receive phone numbers, email addresses, or Social Security numbers,” Ferraz said. “Instead, we receive an anonymized identifier.” That allows data sharing without banks worrying that competitors might use the personal information for advertising or marketing. “Finding a way to enable them to share data privately was the big enabler for this product,” he said. “It’s all about the device, its location behavior, and the anonymized identifier associated with the account.”
The solution connects payment activity and real-world relationships across participating financial institutions, “helping reveal suspicious networks and coordinated activity that individual banks may not be able to identify on their own.” “We already have three banks using this solution,” Ferraz said. The next step, he said, is to gain volume and build a broader network: “We want to establish a consortium of banks that can use this data and share intelligence with one another.” He said the company is seeing a very high level of interest, including from some of the world’s largest financial institutions.
Asked whether regulators may raise privacy concerns given the solution’s reliance on location data, Ferraz said the data is only used for security and fraud prevention purposes. “Precise location information is collected only with the user’s consent,” he said. Among the leading financial institutions Incognia works with, approximately 85 to 95 percent of customers share precise location data. But how the bank communicates this to the end user is extremely important. “Banks that don’t explain why they want to collect location data usually see very low adoption rates,” Ferraz said. Banks that clearly explain the data will be used for account protection, fraud prevention and scam prevention see opt-in rates of 85 to 95 percent.
Article Topics
behavioral biometrics | device fingerprinting | fraud prevention | Incognia | location authentication







Comments