Infostealer
Malware built to silently steal sensitive information from an infected device: passwords, cookies, API tokens, SSH keys, crypto wallets.

An infostealer (information stealer) is a type of malware designed to silently collect sensitive data on an infected device and exfiltrate it to an attacker. Its main targets are passwords saved in browsers, session cookies, authentication tokens (API, SSH), cryptocurrency wallets and payment card data. Stolen credentials accounted for 30% of initial attack vectors in 2024 (IBM X-Force), and SpyCloud recaptured 1.38 billion passwords stolen by infostealers over the same period.
Technical definition
Collection mechanisms
Infostealers target data stored locally by applications. Chromium browsers (Chrome, Edge, Brave) encrypt passwords with DPAPI (Windows) or the Keychain (macOS), but a process running in the user's session can decrypt them. Session cookies are especially valuable: they let an attacker log in to a service without a password or MFA (session hijacking).
Collection techniques include reading the browsers' SQLite databases (Login Data, Cookies), extracting decryption keys from the browser profile, scanning well-known files (.ssh/, .env, credentials.json, crypto wallets), capturing memory (credentials in plain text), and keylogging (recording keystrokes). The collected data is compressed and sent to a C2 (Command and Control) server in a single HTTPS request to minimize detection.
Distribution and campaigns
Infostealer distribution channels evolve with the tech ecosystem. Traditional methods include phishing (email with an attachment), fake software (cracks, keygens) and malicious ads (malvertising). Emerging channels exploit extension marketplaces: malicious npm/PyPI packages (supply chain attacks) and AI agent extensions (ClawHub skills for OpenClaw).
The ClawHavoc campaign, which hit OpenClaw users on macOS, illustrates this new channel: seemingly legitimate skills (productivity, automation) carried an infostealer targeting crypto wallets, SSH keys and browser passwords.
Current challenges
Infostealers-as-a-Service
The MaaS (Malware-as-a-Service) model makes infostealers widely accessible: operators rent their malware to affiliates for $100-300 a month, complete with an admin panel, technical support and anti-detection updates. RedLine, Raccoon Stealer and Lumma are distributed via Telegram and darknet forums. This industrialized model has multiplied the volume of stolen credentials.
A new channel: AI agents
Personal AI agents (OpenClaw, AI browser extensions) create an entirely new attack surface for infostealers. Unlike classic browser extensions (which are sandboxed), AI agents often have access to the file system, the terminal and the network. A malicious AI agent extension can reach data that classic web malware cannot: SSH keys, server configurations, environment variables holding secrets. In technical due diligence, the security of the extension marketplace is a critical criterion.
Standards and Specifications
MITRE ATT&CK - Credential Access
Classification of credential theft techniques in the ATT&CK matrix
NIST SP 800-63B
Guidelines on digital authentication and credential protection
Frequently Asked Questions
How does an infostealer work?
An infostealer runs silently on the victim's device and collects sensitive data stored locally. Its main targets are passwords saved in browsers (Chrome, Firefox, Edge), session cookies (which make it possible to bypass two-factor authentication), API tokens and SSH keys, cryptocurrency wallets, and autofill data (credit card numbers, addresses).
The stolen data is exfiltrated to a command and control (C2) server, then resold on darknet markets (Genesis Market, Russian Market). A full credential bundle (email, password, cookies) sells for $10-50. The most widespread infostealers in 2024 were RedLine, Raccoon, Vidar and Lumma.
How can you protect yourself against infostealers?
Prevention relies on several layers: never install extensions or software from unverified sources (such as the malicious skills on OpenClaw's ClawHub), keep your system and browser up to date, use a password manager (credentials are encrypted, unlike browser storage), and turn on hardware authentication (FIDO2/WebAuthn keys, which resist cookie theft).
Companies deploy EDR (Endpoint Detection and Response) solutions that detect typical infostealer behavior: access to browser profile files, reads of credential databases, communication with known C2 servers. Credential leak monitoring (SpyCloud, Hudson Rock, Have I Been Pwned) makes it possible to detect a compromise quickly.
What is the link between infostealers and AI agents?
AI agents like OpenClaw create a new distribution channel for infostealers. Extensions (skills) from a community marketplace can carry malicious exfiltration code. The ClawHavoc campaign distributed a macOS infostealer through hundreds of malicious ClawHub skills, stealing SSH keys, API tokens and browser passwords.
The risk is amplified by the fact that AI agents have access to the file system and the terminal: an infostealer distributed through an agent extension has more permissions than classic web malware. In due diligence, an unmoderated extension marketplace is a major security red flag.