New exploit targeting Web3 professionals. Attackers pose as recruiters. Victim installs 'Relay' AI meeting software. Real malware. SlowMist confirms. Wallet data stolen. Browser credentials exfiltrated. Telegram sessions hijacked. Audit passed. Trust failed.
Beacon chain stable. Fragility remains.
The timing isn't random. Bull market 2025. Hiring spree across crypto firms. Scammers know the pattern. LinkedIn flooded with fake profiles. Recruiters offer remote gigs. The bait? An AI-powered interview tool. The hook? Relay. The catch? It's a cross-platform info-stealer.
Context: every hiring manager is desperate for talent. Speed trumps verification. That's the opening. Attackers spent time studying Web3 workflows. They know we run hot wallets, store private keys in keychains, keep Telegram open for deal flow. They built for that.
Core: Technical dissection.
SlowMist's sample analysis reveals a mature payload. Two builds: macOS .dmg, Windows .exe. Both signed with stolen or self-signed certs — the code isn't the vulnerability, the trust is. The installer runs a fake meeting UI. Meanwhile, backdoors execute:
- Browser credential theft: Chrome, Firefox, Brave. Saved passwords, autofill data, session cookies.
- Crypto wallet extraction: MetaMask, Phantom, Keplar — any extension or desktop wallet with stored keys.
- Keychain dump (macOS): iCloud, SSH keys, app-specific passwords.
- Telegram session cloning:
.tdesktopfolder,tdatadirectory. Full access to chats without MFA.
Based on my audit experience during the Ethereum 2.0 Beacon Chain slashing conditions, I recognize this pattern. Attackers don't break cryptography. They break human behavior. The Relay malware is no exception. It doesn't exploit a zero-day. It exploits the one thing we rush: speed into a recruiter's link.
Data flow:
Victim downloads → installs → fake UI runs → malware persists → data exfiltration to C2 (SlowMist mapped IPs) → attackers sell credentials or drain wallets.
Cross-platform capability means no safe OS. Windows users lose hot wallets. macOS users lose keychain and Telegram sessions. The attack surface is complete.
Contrarian angle: The irony of AI trust.
We're in a bull market. Everyone is building AI agents for trading, analysis, hiring. The narrative: AI will make crypto faster, smarter, safer. This attack turns that narrative on its head. The promise of AI-streamlined hiring is the vector. The attacker didn't need to hack a smart contract. They didn't need to exploit a bridge. They used the industry's own hype against it.
SlowMist's disclosure is critical. But here's the unreported angle: the real damage isn't the stolen wallets. It's the stolen Telegram sessions. With Telegram access, attackers can impersonate the victim within their company, project groups, deal channels. They can send malicious links to other team members — a second-order attack that spreads faster than the first. The initial victim might never know until coworkers report suspicious messages.
During DeFi Summer, I created a yield optimization model to calculate true APY after gas. I saw how incentives mask reality. This attack does the same: the incentive of a remote job masks the cost of trust. No one checks the hash of the installer before clicking. No one scans the code. They just want the interview.
Takeaway: The next wave of defense.
Expect more of these. The playbook is written. Copycats will modify the payload, rebrand the tool, target different roles (community managers, developers, researchers). What's the fix?
- Hardware wallets only for signing. Never store private keys on devices used for communication.
- Dedicated machines for interviews. Use VMs or sandboxes. Treat every recruiter as hostile until proven otherwise.
- SlowMist's IOC list should be deployed by every crypto company's security team.
But the deeper problem remains. Web3 hiring lacks a trust framework. We have exchange reserve proofs, on-chain audits, but no standard for verifying a recruiter's identity. Until we build that, every remote job offer is a potential exploit.
Question for founders: how much talent will you lose because your hiring process is indistinguishable from a scam?