What is NFC Relay malware and why is it a threat to agentic commerce?

NFC relay malware is a new and rapidly growing form of fraud that increasingly is being used to attack mobile payments – typically hijacking live NFC payments at a point of sale, but also in compromising and sharing tokens that render ‘secure’ mobile payments anything but.

As the world moves towards agentic commerce – where AI chooses and pays for goods and services online – the threat of payment fraud has taken a leap into the unknown. How, for example, would a consumer know that their AI payment has been hijacked? How, for that matter, would the AI itself or the merchant?

The problem is a big one and, as how people shop changes rapidly, NFC relay attacks are only going to get bigger and more pronounced. So, what do telemedia players need to know, what impact will there be on agentic commerce and what can be done about it?

What is NFC relay malware?

NFC relay malware enables criminals to perform “ghost tap” attacks by intercepting and forwarding contactless payment data in real-time

. By exploiting the near-field communication (NFC) protocol, this malware allows fraudsters to conduct unauthorized transactions at a distance, making it appear to merchants and banks as a legitimate tap-to-pay purchase from a victim’s mobile device.

This is a step up from traditional malware that steals credentials, because it directly manipulates the live payment process. The malicious software turns a compromised phone into a relay proxy, which can “borrow” a victim’s payment token and pass it along to a distant accomplice, who then completes the transaction.

A typical NFC relay attack, such as those executed by the “SuperCard X” or “Ghost Tapping” malware, involves the following stages:

1. Victim infection – The scam begins by infecting the victim’s smartphone, often through social engineering methods like phishing text messages or fraudulent phone calls. The victim is tricked into installing a malicious app disguised as a security tool or a legitimate utility. This malware typically requests low-level permissions, making it difficult for standard antivirus software to detect.
2. Payment data acquisition – Using social engineering, fraudsters may instruct the victim to tap their physical credit card against their now-infected phone for “verification”. The malware then uses the phone’s NFC capabilities to read and steal the card’s tokenized payment data from the chip. Some advanced malware can also leverage stolen online credentials and one-time passwords (OTPs) to provision the victim’s card onto a mobile wallet controlled by the criminals.
3. Real-time relay – The attack uses a relay of at least two devices:

The victim’s compromised phone: The malware on this device intercepts the payment information when it is tapped or provisioned and forwards it to the attackers.

The “mule’s” phone: An accomplice, or “mule,” is in a retail store or at an ATM with a second device that has the capability to receive the relayed signal and emulate the victim’s card.

The two phones use a server, often managed on a platform like Telegram, to tunnel the NFC traffic in real-time, effectively extending the short range of NFC over long distances.

4. Unauthorized transaction – With the relayed payment token, the accomplice can perform a contactless tap at a point-of-sale (POS) terminal. The transaction appears legitimate to the merchant’s system because it uses valid, cryptographically signed credentials from the victim’s card. After the purchase, criminals quickly resell the stolen goods, like electronics or gift cards, for cash or cryptocurrency.

How to protect against NFC relay attacks

• Be wary of app installations: Never download apps from suspicious links in text messages, emails, or websites. Stick to official app stores like Google Play, which employs app-scanning measures.
• Enable multi-factor authentication: Use biometrics like fingerprint or face ID to secure your mobile wallet transactions.
• Review app permissions: Check the permissions of your apps and question any app that requests access to NFC features without a clear reason.
• Monitor financial activity: Set up real-time transaction notifications from your bank to be alerted of any unusual payments.
• Keep software updated: Regularly install system and security updates for your phone to patch known vulnerabilities.

Implications for agentic commerce

NFC relay malware attacks pose a significant threat to Agentic Commerce by undermining the foundational principles of user trust, security, and the efficiency that autonomous AI agents promise. These attacks exploit the very technology that enables seamless mobile payments, making it difficult for automated systems to distinguish between legitimate and fraudulent transactions.

The autonomous nature of Agentic Commerce makes it especially vulnerable to NFC relay attacks, creating several critical implications:

Erosion of trust:

• Systemic trust failure: Agentic Commerce relies on a high degree of trust that the AI agent will act in the user’s best interest. If an autonomous payment is proven to be fraudulent, it could shatter consumer confidence not just in a single transaction but in the entire concept of delegating financial decisions to AI.
• Difficulty with accountability: When fraud occurs via an AI agent, it becomes complex to determine who is at fault. Is it the user who installed the malware, the AI platform, the payment network, or the merchant? The ambiguity weakens the trust required for widespread adoption.

Detection and prevention challenges:

• Bypassing conventional fraud detection: From the perspective of a payment network, an NFC relay attack can appear as a legitimate, tokenized, in-person transaction. Traditional fraud detection systems, which look for simple anomalies, may fail because the payment request carries valid cryptographic data.
• Lack of contextual signals: Agentic transactions are designed to operate with minimal human involvement. This removes critical real-time behavioural signals, such as user location and device activity, that human-centric fraud detection relies on.
• Low-value, high-volume fraud: Attackers can use NFC relay malware to conduct numerous small, low-value fraudulent transactions that fly under the radar of typical anti-fraud measures that focus on high-ticket purchases. For an autonomous agent making frequent purchases, these minor discrepancies could aggregate into significant losses.

Disruption of autonomous processes:

• Compromised payment mandates: In Agentic Commerce, AI agents are given mandates to purchase goods based on a customer’s specific request. An NFC relay attack could trick a legitimate-looking agent into approving a fraudulent transaction, disrupting the core function of programmatic purchasing.
• Poisoning the agent’s data: Since agentic AI learns from its interactions, a successful fraud scheme could “poison” the agent’s learned behaviour. The agent might begin to mistakenly flag legitimate user actions as suspicious or, worse, learn to process fraudulent patterns as normal.
• Regulatory uncertainty: The emergence of new attack vectors like NFC relay malware in an AI-driven economy will require new regulatory scrutiny. Governments and payment regulators will need to address questions of liability and risk, potentially slowing the development and deployment of agentic systems while security standards catch up.

Potential mitigation strategies

To counter these threats and secure the future of Agentic Commerce, several measures are being pursued:

• Enhanced fraud detection: Fraud detection systems must evolve to use multi-layered models that incorporate factors like real-time device integrity, geolocation, and user-behavioural monitoring, rather than relying on a single data point.
• Advanced tokenization: Using more dynamic and context-aware payment tokens can help combat replay and relay attacks. Tokenization that includes more granular data, such as a device fingerprint, can add more robust validation.
• Device attestation: Implementing stronger checks to verify the integrity and security of the device initiating the transaction. This could involve cryptographically verifying that the request is coming from a trusted, uncompromised environment.
• Layered authorization: Building in additional checks for the agent itself. For instance, requiring two-factor or biometric confirmation for transactions above a certain monetary threshold can limit the impact of an exploited agent.