
LuckyTaoremWhat Is the Pass‑ta‑key Attack? Last week, Palo Alto Networks researcher Arie Olshtein...
Last week, Palo Alto Networks researcher Arie Olshtein published a proof‑of‑concept that challenges a core security belief: passkeys are stored only inside the Trusted Platform Manager (TPM), a hardware enclave designed to be tamper‑proof. The new technique, dubbed Pass‑ta‑key (a playful blend of “passkey” and “pass the key”), shows that when a Windows machine is compromised by malware, an attacker can extract every passkey stored in the Google Password Manager (GPM) app for Windows.
The attack does not invent a brand‑new cryptographic flaw; instead, it repurposes classic credential‑stealing malware techniques to target the data that the GPM sync client keeps on the local filesystem. The novelty lies in the fact that the community has long treated passkeys as “hardware‑only” secrets, assuming the TPM would shield them from any software‑level compromise.
Google’s Password Manager for Windows maintains a local encrypted vault that mirrors the user’s cloud‑based passkey store. While the encryption keys are derived from TPM‑protected material, the encrypted blob itself resides on the file system in a location accessible to the user profile. This design enables offline authentication and faster login, but it also creates a “soft” attack surface.
Pass‑ta‑key assumes the attacker already has persistence on the victim machine. Typical infection vectors include:
Once the malicious code runs with user‑level privileges (or higher), it can read the GPM vault file, locate the TPM‑derived decryption key in memory, and decrypt the passkeys.
🔹 ------
• Action: --------
• Technical Detail: -------------------
🔹 *A*
• Action: Locate the GPM vault
• Technical Detail: Usually stored under %APPDATA%\Google\PasswordManager\ as vault.dat.
🔹 *B*
• Action: Dump TPM‑derived key material
• Technical Detail: The key is loaded into process memory when the GPM client starts; malware can use Windows API calls like ReadProcessMemory to capture it.
🔹 *C*
• Action: Decrypt the vault
• Technical Detail: The vault uses AES‑256‑GCM; with the extracted key, the attacker can decrypt the entire file.
🔹 *D*
• Action: Export passkeys
• Technical Detail: Decrypted data contains WebAuthn credential IDs, public keys, and encrypted private keys. The attacker can re‑wrap the private keys for use in a rogue authenticator.
The steps above mirror classic credential‑stealing tools that target password databases (e.g., lsass.exe dumps, Chrome password extraction). Pass‑ta‑key simply swaps the target from passwords to passkeys, proving that the “new” attack surface is really a re‑contextualization of known methods.
The TPM is a hardware root of trust, but it does not magically protect data that is later written to disk. Passkeys rely on the TPM for key generation and attestation, yet the encrypted credential store lives outside the chip. This separation creates a gap that attackers can exploit once they have code execution on the host.
Many end users and even security professionals have promoted passkeys as a “password‑free, phishing‑proof” solution. The Pass‑ta‑key research reminds us that phishing‑proof only holds when the authentication flow remains intact and the device itself is uncompromised. If malware can harvest the private keys, the attacker can replay or forge authentication requests.
Companies that have rolled out passkey‑only login for internal SaaS tools are now revisiting their zero‑trust models. The consensus is that passkeys should be one factor in a layered approach that still includes device posture checks, behavioral analytics, and continuous authentication.
vault.dat file outside of the legitimate Google client.The Pass‑ta‑key discovery is a wake‑up call, not a death knell. The underlying cryptographic primitives of passkeys (WebAuthn, FIDO2) remain robust. What is at stake is the operational security of the surrounding software stack.
Term*: The industry will likely see a shift toward **passkey portability standards* that balance security with user convenience, such as encrypted backups that require multi-factor authentication (MFA) for restoration.
Passkeys were designed to eliminate passwords, but the Pass-ta-key attack underscores a critical truth: no authentication method is immune to endpoint compromise. The attack doesn’t break the cryptographic guarantees of passkeys—it exploits the gap between hardware-backed security and real-world software implementations. This tension isn’t unique to passkeys; it mirrors the evolution of other authentication technologies, from smart cards to hardware tokens, where the weakest link often lies in how secrets are stored and accessed.
Passkeys are following a similar trajectory. The TPM provides a strong foundation, but the ecosystem around it—sync clients, local storage, and memory management—introduces vulnerabilities that attackers will inevitably target.
The Pass-ta-key attack reinforces the principles of zero trust architecture (ZTA). In a zero-trust model:
For passkeys, this means:
The Pass-ta-key research has sparked conversations among industry consortia, including:
Read the full breakdown originally published at https://ltdeveloperblogs.github.io/posts/new-pass-ta-key-attack-reveals-all-the-things-we-didnt-know-about-passkeys/