Stop believing your hardware wallet makes you immune. A new class of social engineering malware just bypassed the last line of defense — your trust in a job offer.
On July 29, 2025, SlowMist published an analysis of a targeted attack against Web3 professionals. The delivery mechanism: a fake AI meeting app called "Relay," disguised as a recruiter's screening tool. The payload: a cross-platform information stealer that harvests browser credentials, private keys, Telegram sessions, and macOS Keychain data. The attack chain is clean, surgical, and terrifyingly effective.
I've spent the last seven years auditing liquidity, not malware. But when you manage a digital asset fund, you learn to read attack vectors the same way you read yield curves — as signals of where value is most vulnerable. This one signals a structural shift. The attackers are not script kiddies. They understand the Web3 hiring pipeline, the desperation in a sideways market, and the single point of failure that no cold storage can protect: human trust.
Context: The Macro Background of a Phishing Wave
The current market is consolidating. Bitcoin oscillates in a range, altcoins bleed liquidity, and the narrative around AI tools has created a fertile ground for exploitation. In a bull market, frauds are usually about greed — rug pulls, Ponzi schemes. In a sideways market, frauds shift to hope. People want jobs. Teams want to hire cheaply. The attackers exploit this friction.
The "Relay" app is not a real product. It's a custom-crafted malware binary that targets both macOS and Windows. According to SlowMist's samples, the malware performs extensive reconnaissance: it scrapes browser data (cookies, autofill, saved passwords), dumps cryptocurrency wallet extensions (MetaMask, Phantom, Keplr — any Chromium-based extension is at risk), and exfiltrates Telegram session files. The Telegram angle is critical — once an attacker has your session, they can impersonate you to your network, amplifying the attack into a multi-stage compromise.
This is not a generic phishing link. It's a targeted social engineering operation that requires the attacker to build a credible recruiter persona, often on LinkedIn or Telegram, engage in conversation, and then convince the target to download and run an unsigned binary. The investment in time and effort indicates high-value targets: Web3 engineers, protocol founders, and fund managers — anyone with access to significant crypto assets or signing authority.
Based on my fund's due diligence on security audits during the 2021 NFT cycle, I learned that the most sophisticated attacks exploit peripheral trust. When I evaluated the Ronin bridge security before its hack, I flagged the reliance on a small set of validators and the lack of session isolation for operational staff. That same principle applies here: the attacker bypasses technical defenses by attacking the human operating system.
Core: Technical Anatomy of the Attack
Let me walk you through why this malware is dangerous — not because it's novel, but because it targets the one thing every Web3 professional does daily: communicate with strangers about opportunities.
1. Delivery Mechanism The fake recruiter reaches out via LinkedIn or Telegram. The conversation is standard: company pitch, role description, request for a screening call. Then they ask you to download a meeting tool. "We use Relay for AI-powered interview analysis. Please install before the call." The binary is hosted on a domain that mimics a legitimate startup. The code is signed with a dodgy certificate, but most users ignore the warning.
2. Cross-Platform Payload The malware is written in a cross-platform framework (likely Go or Electron, based on the binary size). It installs on macOS with a fake bundle identifier and on Windows with a misleading executable name. Once run, it immediately copies itself to persistence locations: - macOS: LaunchAgents or LaunchDaemons - Windows: Registry Run keys or Scheduled Tasks
3. Data Exfiltration Scope The malware scrapes: - Browser credential stores (Chrome, Brave, Firefox, Safari) - Crypto wallet extension private keys (some wallets store keys unencrypted in local storage — a known vulnerability that attackers exploit) - Telegram session files (allows complete impersonation) - macOS Keychain (passwords, tokens, SSH keys) - Environment variables and clipboard history (many developers paste private keys into terminals)
This is not a remote access trojan (RAT) that waits for commands. It's a pure stealer: gather everything, compress, and send to a C2 server via encrypted HTTPS. The attackers don't need persistence after exfiltration, but they maintain it to capture future activity.
4. The Telegram Session Attack This is the most underappreciated risk. Telegram stores session tokens in a local file (tdata folder on Windows, ~/Library/Group Containers/ on macOS). If an attacker obtains this file, they can clone your session without needing your phone number or 2FA. They can then read all your chats, send messages as you, and access your contact list. For a Web3 professional, this means the attacker now has access to your private founder groups, project channels, and negotiation threads. They can pivot to anyone you trust.
I've seen this before in the 2022 Discord token hijacks, but those required the user to authenticate a malicious OAuth app. This is worse — no authentication needed. Just file access.
Contrarian: The Decoupling Thesis — Why Cold Storage Is Not Enough
Conventional wisdom says: use a hardware wallet, never store private keys on a hot device, and you're safe. This attack disproves that assumption in two ways.
First, many Web3 professionals still use browser-based wallets for daily operations. A hardware wallet doesn't protect you when you sign a transaction from MetaMask on a compromised machine. The attacker can wait until you interact with a dApp and capture the signed transaction or swap the recipient address via malicious browser extension injection.
Second, the attack on Telegram sessions creates a secondary threat that no cold wallet can mitigate. Even if your funds are safe in cold storage, your reputation and access to project communications are not. An attacker can join your protocol's governance call, copy sensitive documents, or draft fake proposals.
The real decoupling is this: security is not a feature of your hardware; it's a process of your behavior. The market is starting to realize that the marginal return on better encryption is diminishing compared to the return on better operational security. This attack will accelerate the adoption of zero-trust architectures in Web3 companies.
Experience Signal: The Ronin Bridge Lesson
When I pivoted our fund away from NFT PFPs into blockchain gaming infrastructure in early 2022, I personally audited the security setup of Axie Infinity's Ronin bridge. I noticed a gap: the bridge operators used shared Telegram groups for emergency alerts. I recommended a separate Signal channel and hardware-backed signing. The team didn't implement it before the hack. The hackers got in via a compromised private key from a former employee's session. The same vector — session theft — is now being weaponized at scale.
If you think your cold wallet protects you, you're missing the point. The asset that matters most in a sideways market is your network. And your network is only as secure as your Telegram session.
Takeaway: Liquidity Vanishes Faster Than Hype
The headline is the attack, but the narrative is the lesson. Security in Web3 has always been reactive. This attack is a signal that the next phase of adversarial innovation will target the human layer — not the smart contract.
Don't trust the yield; audit the source. And right now, the source is a job offer. Verify the recruiter through independent channels. Never run unsigned binaries. Use a dedicated laptop or VM for any interview that requires software installation. Treat your Telegram session like your master key — because it is.
Liquidity vanishes faster than hype. But trust, once stolen, doesn't come back. The question for every Web3 professional reading this: Are you ready to trust a protocol that can't even verify its own recruiters?