What triggered the discovery
Microsoft Threat Intelligence researchers identified a campaign where compromised websites embed scripts that read malicious instructions directly from BNB Chain transactions. The attackers write encoded PowerShell commands to the blockchain, then have the website fetch and decode them at runtime. Visitors encounter a fake CAPTCHA that copies a one-liner to their clipboard and instructs them to press Win+R, paste, and hit Enter — executing the payload on their Windows machine.
This discovery highlights a growing trend where threat actors leverage programmable blockchains not for financial theft but as resilient infrastructure for malware delivery. Unlike traditional command-and-control servers that can be seized or sinkholed, blockchain-based instructions persist across a distributed network of validators, making takedown efforts significantly more complex and time-consuming.
How the infection chain works
The fake CAPTCHA mimics a "I'm not a robot" check. When clicked, it runs `navigator.clipboard.writeText()` to place a PowerShell command on the clipboard. A follow-up overlay tells the user to open the Run dialog and paste. The command downloads and executes a second-stage script from an attacker-controlled server. Because the initial payload lives on-chain, the operators can update it without touching the compromised web server, evading traditional file-based detection.
The technique exploits a fundamental trust assumption: users expect CAPTCHAs to verify humanity, not to initiate system-level commands. By copying the payload to the clipboard rather than executing it directly in the browser, the attack sidesteps browser sandbox protections and shifts the execution context to the operating system itself — where PowerShell runs with the user's full permissions.
| BNB price | $593.48 |
|---|---|
| 24h change | -0.20% |
| 7d change | +1.20% |
| 24h volume | $559.4M |
| Market cap | $79.03B |
| Rank | #4 |
BNB Chain's low transaction fees and fast block times make it economical for threat actors to treat the ledger as a mutable configuration store. The chain processes roughly 3-5 million transactions daily, giving ample cover for the small number of writes this campaign requires. At a cost of fractions of a cent per transaction, attackers can update their payloads frequently without incurring meaningful expense, while the high volume of legitimate traffic provides effective camouflage.
Why the timing matters
The campaign coincides with a period of elevated on-chain activity across major networks. Total crypto market cap sits at $2.30 trillion with 24-hour volume of $53.4 billion. BTC dominance holds at 56.8%. Researchers note that attackers increasingly favor programmable chains for command-and-control because takedown requires coordination with validators or RPC providers — a slower process than seizing a domain or server.
This shift reflects a broader evolution in cybercrime infrastructure. As defenders improve at detecting and blocking traditional C2 channels — domain generation algorithms, compromised web servers, and cloud-hosted relays — attackers are migrating to infrastructure that is inherently resistant to unilateral takedown. Blockchain networks, by design, lack a central authority that can arbitrarily remove data, forcing defenders into a coordination game across dozens of independent validator operators and RPC providers.
What happens next
Microsoft has shared indicators of compromise with security vendors and blockchain analytics firms. RPC endpoints and block explorers can flag the malicious contract addresses. Users should treat any CAPTCHA that asks for a Run-dialog paste as malicious; legitimate CAPTCHAs never require clipboard interaction outside the browser. Enterprises should block outbound PowerShell execution from user workstations where feasible.
The response underscores a critical gap in current defenses: most endpoint protection platforms monitor for malicious file downloads and script execution, but few inspect clipboard contents or validate the legitimacy of clipboard-mediated command chains. As this technique proves effective, it will likely be adopted by other threat groups and adapted for macOS and Linux targets using equivalent shell execution vectors.