July 29, 2025 – SlowMist’s threat intelligence team flagged a new malware variant masquerading as an AI meeting tool named "Relay." The binary is cross-platform, targeting macOS and Windows. Its payload: credential theft from browsers, crypto wallets, keychain services, and Telegram sessions.
The attack vector is a hiring scam. Fake recruiters, often claiming to represent legitimate Web3 firms, invite candidates to install "Relay" for a technical interview. Once installed, the malware exfiltrates all sensitive data. The target profile? Web3 professionals—developers, researchers, marketers—anyone with access to private keys or project infrastructure.
This is not a new technique. Social engineering has been the crypto industry’s Achilles’ heel since 2017. But the scale and sophistication here reveal a systemic vulnerability: the industry’s reliance on trust in professional interactions, despite its core promise of trustless systems.
Structure beats speculation every time. That mantra guided me through the ICO mania, the DeFi summer, and the NFT crash. It applies here too. The speculation is that AI tools make remote work seamless. The structure is that every new tool introduces a new attack surface. SlowMist’s analysis provides the blueprint: a fake resume, a cloned LinkedIn profile, a malicious .dmg file, and a silent data drain. No zero-day exploit. No smart contract flaw. Just a human clicking "Install."
2017 called. It wants its lessons back. That year, I dissected over 500 Ethereum whitepapers. Eighty-five percent had no viable roadmap. The pattern repeats: hype around a new technology (AI meeting software) masks the lack of security rigor. The attackers are not exploiting code—they’re exploiting the narrative that "AI will revolutionize hiring."
Hook: The Binary That Exposed a Broken Trust Model
On July 28, 2025, a security researcher in Southeast Asia received a LinkedIn message from "Sarah Chen," a supposed talent acquisition specialist for a well-known Layer-2 scaling project. The message was polished, referencing recent blog posts about zk-rollups. The candidate was asked to download "Relay"—a lightweight AI transcription tool—for a technical screen. The file was a signed macOS application with a valid developer certificate, likely stolen or created via a compromised account. Within hours, the researcher’s crypto wallet was drained, and his Telegram session hijacked.
SlowMist captured the binary. The code obfuscation was minimal—the attackers didn’t need sophistication. The social engineering was the real payload. The malware’s operation was simple: inject into browser processes, dump saved credentials, scan for common wallet directories, and exfiltrate via encrypted WebSocket. The Windows variant used DLL sideloading; the macOS variant exploited a code signing bypass that Apple patched two years ago—but many users skip updates.
This incident is not isolated. Over the past three months, at least 12 similar attempts have been reported on crypto Twitter under the hashtag #HiringScam. The common denominator: fake recruiters using AI-generated headshots and cloned company logos. The attack is scalable because it leverages the industry’s hunger for talent and its tolerance of remote, asynchronous hiring.
Context: The Historical Precedent of Trust Exploitation
The crypto industry was built on the idea of eliminating intermediaries—banks, escrow agents, rent-seeking platforms. Yet hiring remains a deeply centralized, trust-based process. Recruiters hold gatekeeper power. Job postings are verified by reputation, not cryptographic proof. A fake LinkedIn profile with a few connections and a custom URL can fool even seasoned founders.
In 2017, I saw ICO teams raise millions with no code, just a whitepaper and a Telegram group. The underlying narrative was that technology would solve trust. The 2017 crash revealed the lie: trust in technology without trust in execution is just speculation. The same pattern repeats here. The narrative is that AI tools will streamline hiring. The reality is that these tools create an unverified trust layer that attackers can inject their payload into.
Consider the parallels with the 2021 NFT social engineering wave. Attackers posed as collectors, artists, or curators to gain access to wallets. The method was identical: establish credibility through fake social proof, then request a "tool" (e.g., a wallet connector that was a drainer). The crypto community responded with education—but education alone cannot stop a determined attacker. The difference now is the target demographic: Web3 professionals with access to private keys, infrastructure secrets, and protocol administration rights.
2017 called. It wants its lessons back. That year, the lesson was that code is not trust. This year, the lesson is that identity is not trust. A LinkedIn profile with 500 connections, a company logo, and a history of Web2 job titles is not sufficient proof of authenticity. The industry needs a new layer: verifiable credentials for professional interactions.
Core: The Anatomy of the Relay Attack
The malware sample (SHA256: a3f2c9…e1b4) was hosted on a domain that mimicked a real AI meeting startup. The domain was registered two weeks before the attack, with privacy protection enabled. The HTTPS certificate was issued by Let’s Encrypt, which is free and automated—no barrier to abuse. The binary itself was a standard Electron application with a native addon that performed the theft.
Data Exfiltration Scope: - Browser Credentials: Chrome, Firefox, Brave, and Opera saved passwords from under the hood (the attacker could collect logins to email, GitHub, CI/CD pipelines, and cloud consoles). - Crypto Wallets: Scanned for common paths: ~/.ethereum, ~/Library/Ethereum, ~/.keystore, and browser extensions like MetaMask, Phantom, and Keplr. The malware copied keyfiles and captured running wallet passwords via keylogging. - System Keychain (macOS): Dumped the user’s login keychain via a known vulnerability in the security command-line tool. This grants access to SSH keys, application passwords, and any secrets stored by the user. - Telegram Sessions: The malware located the Telegram desktop app’s session cache (tdata folder) and exfiltrated it. This allows the attacker to impersonate the victim in group chats, access private channels, and reset passwords for services linked to Telegram.
Attack Flow: 1. Fake recruiter establishes contact via LinkedIn, Discord, or Telegram. 2. After initial rapport, the recruiter sends a calendar invite with a link to download "Relay." 3. Victim downloads and runs the installer. The native addon requests full disk access (macOS) or Administrator (Windows). 4. Malware executes in memory, avoiding disk-level detection. 5. Stolen data is encrypted and sent to a command-and-control server (C2) via WebSocket. The C2 domain rotates every 24 hours.
Why is this effective? - The victim is expecting third-party software for the interview. The trust is already established by the recruiter’s narrative. - The malware mimics a legitimate Electron app; the visual interface is identical to the real "Relay" app (which exists but is obscure). - Code signing certificates can be acquired by paying a certificate authority ($99 for Apple Developer, $199 for Microsoft). Attackers use stolen credit cards or compromised accounts. - Antivirus engines flagged the binary as a generic Trojan only after SlowMist’s report. Initial detection rates were below 10%.
Personal Experience: During my 2020 DeFi consulting work, I audited several projects where the founder’s computer was compromised via a fake Slack integration. The attacker gained access to the GitHub repository and pushed a backdoor to the yield aggregator’s code. It was caught in a code review, but it taught me that the weakest link is always the human terminal. Based on that experience, I now advise clients to use air-gapped machines for any wallet operations. But that advice is impossible for a job interview. The attack surface is unavoidable unless the industry rethinks the hiring process itself.
The Edge: Why This Attack Is Different from Past Scams Prior hiring scams typically involved fake job offers that required candidates to send small amounts of crypto for "processing fees" (a classic advance-fee scam). The Relay attack is a step-change: it delivers actual malware that compromises the entire digital identity of the target. The return on investment for attackers is massive. A single compromised Web3 developer’s laptop can yield access to multiple protocol admin keys, team wallets, and sensitive Telegram groups.
Contrarian: The Real Risk Is Trust, Not Malware
The industry’s immediate reaction will be to install better endpoint protection, use hardware wallets, and educate employees. These are necessary but insufficient. The contrarian perspective: the malware is a symptom of a deeper structural flaw in Web3 professional interactions.
Blind Spot 1: The Myth of Transparent Recruiting Crypto projects pride themselves on transparency: on-chain treasury, public GitHub repos, open roadmaps. Yet hiring is opaque. There is no standard way to verify a recruiter’s affiliation with a project. Most projects do not publish a list of authorized recruiters or use signed messages to verify job offers. Attackers exploit this vacuum. The industry’s narrative of "trust the code, not the person" collapses when the code is the malware and the person is the attacker.
Blind Spot 2: The Failure of Identity Protocols Decentralized identity (DID) and verifiable credentials have been discussed for years. The Relay attack exposes the urgent need for a DID-based hiring standard. Imagine a future where a recruiter sends a verifiable credential signed by the project’s DAO multisig. The candidate’s wallet verifies the signature before downloading any software. This is not a technological challenge—it’s an adoption challenge. The industry prefers shiny new protocols over boring identity infrastructure.
Blind Spot 3: The Unintended Consequences of AI Hype The narrative that "AI will revolutionize X" is being weaponized. "Relay" is an AI meeting tool, and its fake version exploits the trust in AI making interviews easier. The contrarian view: the crypto industry is currently in a period of AI-crypto convergence hype, and this attack is a preview of the next wave of scams that will leverage deepfake audio/video, AI-generated interview questions, and automated social engineering. The underlying skepticism from my experience: every new narrative creates new attack surfaces. In 2017, it was ICO whitepapers. In 2021, it was NFT profile pictures. In 2025, it’s AI meeting apps.
Personal Contrarian Opinion: I have always argued that "delegation makes governance more centralized" in DAOs. The same logic applies here: delegating the verification of a recruiter to a platform like LinkedIn is centralizing trust. The Relay attack shows that this centralized trust is porous. The solution is not to trust the platform but to make verification a cryptographic challenge.
Takeaway: The Next Narrative Is Zero-Trust Hiring
The Relay attack is not the last of its kind. It is the first in a new class of attacks that blend social engineering with AI-powered tools. The next narrative will shift from "AI will streamline hiring" to "verified interactions will secure hiring." I expect to see: - Startups building protocol-level identity verification for hiring. Projects like Ceramic, ENS, and verifiable credentials will gain traction. - Hardware wallet manufacturers integrating professional identity features. Ledger and Trezor have SDKs for signing credentials; they could add a "Safe Interview" mode that runs a sandboxed environment. - Security firms like SlowMist expanding their services to include recruiting security audits—testing a project’s hiring process for vulnerabilities.
Structure beats speculation every time. The speculative narrative is that AI will transform work. The structural reality is that every new tool brings new risks. The market will eventually price in the cost of security failures. Projects that adopt zero-trust hiring now will survive the next bear market; those that don’t will be drained.
2017 called. It wants its lessons back. The lesson is that trust must be built, not assumed. The Relay attack is a reminder that the crypto industry’s greatest vulnerability is not its code, but its people—and the systems that connect them.
The question is: who will build the next layer?