11 Mar Smile ID finds worrying security threats in Africa

Artificial intelligence is fuelling a sharp rise in the scale and sophistication of digital fraud, according to a new report.

The study, titled From Selfies to Signals: Identity Enters the Security Era, was published by Smile ID.

Drawing on data from more than 200 million identity checks conducted in 2025 across 37 industries in over 35 countries, the report concludes that fraudsters are increasingly targeting the infrastructure used to verify identity rather than attempting to manipulate images or documents.

The findings suggest that advances in artificial intelligence have dramatically lowered the cost of producing convincing deepfakes while allowing attackers to operate at a far greater scale.

At the same time, identity systems themselves are becoming integral to digital services ranging from banking to ride-hailing, creating a broader attack surface for criminals.

In 2025, nearly 90 per cent of fraudulent activity blocked by the company was detected through mobile software development kit signals rather than through image analysis alone.

The way identity is verified is now as important as the image presented, the report says. The company recorded more than 100,000 so-called “injection-style” fraud attempts each month linked to emulators, virtual cameras and manipulated device environments designed to bypass verification systems.

Authentication-related fraud has become the dominant threat. According to the research, such attempts now occur more than five times as frequently as fraud during the initial onboarding process, suggesting attackers are increasingly exploiting verified accounts rather than trying to create new ones.

Instead of breaking into systems, criminals are apparently targeting login processes, account recovery tools, device changes and high-value transactions once access has been established.

The report says automated tools powered by artificial intelligence are enabling attackers to reuse biometric data, hijack user sessions and move funds between platforms at speed. As a result, digital fraud is evolving from isolated incidents of document manipulation into organised campaigns by criminal networks exploiting structural weaknesses in mobile-first systems.

Because the company operates as a shared identity infrastructure across multiple industries, it is able to identify patterns of coordinated fraud that may not be visible to individual institutions.

By combining traditional algorithms, controlled capture methods and internally tuned large language models, the company says its systems analyse privacy-preserving metadata to detect coordinated attacks.

Among the report’s key findings is that authentication fraud attempts are now five times more common than those occurring during account creation.

Nearly 90 per cent of fraud blocked in 2025 was triggered by mobile SDK signals, compared with 68 per cent in 2024.

Attempts involving duplicate identities more than doubled year on year and almost tripled the combined totals recorded in 2023 and 2024.

The report also warns that deepfake-driven fraud is increasing across multiple markets as artificial intelligence tools become cheaper and easier to access.

Injection attacks, in which synthetic or pre-recorded media bypass the camera entirely, have emerged as a central threat category, the report says.

Mark Straub, chief executive of Smile ID, said the findings reflected a fundamental shift in how digital fraud operates.

He said: “Fraud is no longer a ‘KYC’ problem – it is a continuous cybersecurity challenge. AI enables fraudsters to operate at unprecedented scale and sophistication. Effective defence now requires network intelligence: By leveraging these privacy-preserving indicators throughout the customer lifecycle, we enable real-time adaptation. Identity has entered the security era, where ecosystem-wide protection is essential to safeguarding the individual.”