Unmasking Cybercrime Strengthening Digital Identity Verification against Deepfakes 2026
Page 13 of 23 · WEF_Unmasking_Cybercrime_Strengthening_Digital_Identity_Verification_against_Deepfakes_2026.pdf
Persistence, stealth
and operational traces
Persistence: Desktop tools commonly persisted via
Windows services, scheduled tasks or installed drivers.
Mobile persistence frequently relied on combination with
root frameworks (e.g. modules that reload on boot). Some
virtualization/container approaches remained active only
while the virtual environment was installed and required
explicit user action to relaunch.
Stealth: Most tools lacked strong stealth and were visible in
task managers, package lists or driver registries. Root/jailbreak
indicators and module identifiers were consistently recoverable.
KYC compatibility and
KYC risk implications
Browser flows versus SDKs: Virtualization and feed
redirection tools were observed to sometimes bypass
browser-based flows (particularly when manual setup for
browser permissions was performed), but these approaches
generally failed against modern SDK-based verification
systems that embed integrity checks and use challenge–
response mechanisms.
Risk to live KYC flows was assessed as conditional:
• Higher risk was observed where tools combined 1) virtual
camera drivers or reliable injection paths, 2) real-time
streaming capability, and 3) low latency with accurate timing.
• Moderate to low risk was attributed to rootless
or offline tools that relied on pre-recorded media,
strict format requirements or manual setup. These
tools could still be repurposed for replay or injection attacks, but the additional steps increased the number
of detectable signals.
Detection and risk mitigation
Detection and risk mitigation efforts were advised to prioritize
the following signals and controls:
1. Hooking framework detection: Flag the presence of Xposed/
LSPosed/Magisk modules and other hooking artefacts.
2. Virtual device enumeration: Monitor for unusual or
duplicate camera device registrations and non-standard
device class IDs.
3. Filesystem indicators: Scan for non-standard media
directories, override files, cloned package names and
container-related package families (e.g. com.fvbox.*).
4. Driver and registry monitoring (desktop): Detect
installation of signed virtual webcam drivers, recently
added registry keys associated with device drivers,
and unexpected Windows services.
5. Timing and challenge correlation: Implement strict
challenge–response timing checks and multi-modal
synchronization (audio versus visual) to increase false
positive resilience and detect timing slippage.
6. Network and process telemetry: Correlate local
upstream streaming patterns with the creation of virtual
camera devices to detect feed injection attempts.
7. SDK integrity: Where feasible, prefer SDK-based
verification modules with embedded integrity checks
over pure browser flows.
Unmasking Cybercrime
13
Ask AI what this page says about a topic: