Join our Newsletter — 33% off our NHI Course
Home› Guides› T1555 Credentials from Password Stores Defensive Guide: Protecting…
Guide Attack Techniques & Malware

T1555 Credentials from Password Stores Defensive Guide: Protecting Browsers, Vaults and Cloud Secrets

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 27 October 2026 13 min read
On this page

Every place built to remember a password for someone, a browser, the operating system, a password manager or a cloud secrets service, is also a single place an attacker can read many passwords at once. MITRE ATT&CK tracks this as T1555 Credentials from Password Stores in the Credential Access tactic (TA0006), with six sub-techniques in v20 covering macOS Keychain, securityd memory, web browsers, Windows Credential Manager, password managers and cloud secrets management stores. Browser credential theft by infostealers is the most common form: ATT&CK links T1555.003 to more groups and software than any other sub-technique here. This guide explains the technique at the ATT&CK description level, then covers the detection strategies MITRE maps, the browser, Windows, macOS and cloud settings that reduce exposure, and how to measure progress. It is for endpoint, identity and cloud security teams and detection engineers.

Key takeaways

  • The store is the target, not the user. Code running as the signed-in user can usually ask the operating system to decrypt that user's saved browser passwords, so a single infostealer run yields every saved login.
  • Machine secrets end up in human stores. API keys, cloud console passwords and service account credentials saved in browsers and personal vaults turn one infected laptop into access for workloads that have no MFA.
  • Cloud secrets managers are password stores too. T1555.006 covers an attacker with a privileged cloud identity simply calling the retrieval API, so least privilege on secret read permissions is the main control.
  • Detection is "who opened the store". MITRE's analytics look for processes other than the owning application reading credential files or memory, and for unusual identities retrieving secrets in bulk.
  • Reduce what is stored. Disable browser password saving on managed devices where an enterprise password manager exists, block network credential caching where not needed, and replace stored secrets with federated or short-lived credentials.

How the attack works

ATT&CK describes T1555 as adversaries searching common password storage locations to obtain user credentials, including password managers and cloud secrets vaults, then using them for lateral movement and access to restricted information. From a defender's view the technique depends on three conditions:

  • Something is stored: a browser, Credential Manager, Keychain or vault holds reusable passwords or secrets.
  • The protection is tied to the user, not the application: many stores decrypt for any process running in the user's session, or for an identity with read permission on the service.
  • The foothold matches: malware running as the user, root on macOS or Linux, or a cloud identity that can call the secret retrieval API.
IDName (ATT&CK v20)StorePlatforms
T1555.001KeychainmacOS Login, System and Local Items (iCloud) keychains under ~/Library/Keychains/ and /Library/Keychains/macOS
T1555.002Securityd MemoryMaster key material held in the memory of the securityd daemon, readable with rootLinux, macOS
T1555.003Credentials from Web BrowsersSaved logins in Chrome, Edge, Firefox and Safari profile databases, protected on Windows by DPAPILinux, macOS, Windows
T1555.004Windows Credential ManagerWeb Credentials and Windows Credentials lockers in .vcrd files with a Policy.vpol key fileWindows
T1555.005Password ManagersThird-party vault databases and the decrypted contents or master password in memoryLinux, macOS, Windows
T1555.006Cloud Secrets Management StoresAWS Secrets Manager, Google Cloud Secret Manager, Azure Key Vault, HashiCorp VaultIaaS

T1555.006 is distinct from T1552.005 Cloud Instance Metadata API: the attacker asks the secrets service directly rather than reading credentials through the metadata endpoint. Related credential access techniques are T1003 OS Credential Dumping (including T1003.001 LSASS Memory) and T1539 Steal Web Session Cookie, which infostealers usually perform in the same run. See the ATT&CK Credential Access Tactic Guide.

Where it shows up

  • Infostealer infections: commodity stealers harvest browser passwords, cookies and autofill on execution and sell the logs. Mandiant linked the 2024 Snowflake campaign to credentials taken by stealers including VIDAR, RISEPRO, REDLINE and LUMMA, some from contractor laptops. See the Snowflake Breach 2024 and the Infostealer Threat Guide.
  • Browser sync to personal accounts: in the Cisco Breach 2022, an employee's corporate password was saved in the browser and synced to a personal Google account; taking over that account gave the attacker the credential, and MFA fatigue did the rest.
  • Human vaults holding machine keys: in the LastPass Breach 2022, a keylogger on an engineer's home computer captured the corporate vault master password, and the vault held the AWS access keys and decryption keys for customer backups.
  • Cloud control planes: ATT&CK links HAFNIUM, Storm-0501 and TeamPCP to T1555.006, and Scattered Spider and LAPSUS$ to T1555.005.
  • macOS fleets: stealers aimed at macOS prompt for the user's password to unlock the Login Keychain. See the Atomic macOS Stealer Guide.

Impact

  • Many accounts from one event: a browser store typically holds dozens of logins, including SaaS admin consoles, VPN portals and cloud consoles.
  • Reuse into privileged access: ATT&CK notes that browser credentials which overlap with privileged accounts, such as domain administrators, significantly further an intrusion.
  • NHI exposure: API keys and service account passwords saved by developers usually have no MFA, no session binding and long lifetimes, so they work from anywhere. See the MITRE ATT&CK T1078 Valid Accounts Guide.
  • Cloud-wide secret disclosure: a single over-privileged role that can list and read all secrets exposes database passwords, signing keys and third-party tokens for every workload.

Detection

ATT&CK v20 maps one detection strategy to the parent and to each sub-technique. The analytics and log sources below are read from the official v20 dataset; see the ATT&CK Detection Strategies and Analytics Guide for how strategies, analytics and data components fit together.

TechniqueDetection strategyBehaviour to detectLog sources named by MITRE
T1555DET0430 Detect Credentials Access from Password StoresNon-standard processes reading credential store files or accessing LSASS, DPAPI or browser databases; bulk secret retrieval in cloudSysmon 1 and 10; Security 4656, 4663, 4670; auditd open/read; macOS unified log; CloudTrail GetSecretValue, Decrypt
T1555.001DET0396 Detect Access to macOS Keychain for Credential TheftThe security utility or osascript run with unusual parents, direct reads of login.keychain-db, unlock attemptsmacOS unified log
T1555.002DET0057 Detect Suspicious Access to securityd Memoryptrace or task_for_pid against securityd; debuggers attached to credential daemons on LinuxmacOS unified log; auditd ptrace and /proc/*/mem reads
T1555.003DET0037 Detect Suspicious Access to Browser Credential StoresProcesses other than the browser reading Login Data or logins.json, followed by CryptUnprotectData calls or browser memory accessSecurity 4656, 4663, 4670 (with object access auditing); Sysmon 1 and 10; auditd file watches
T1555.004DET0134 Detect Suspicious Access to Windows Credential Managervaultcmd.exe or rundll32.exe keymgr.dll launched unexpectedly; reads of .vcrd and .vpol filesSecurity 4688; Sysmon 10 and 15
T1555.005DET0597 Detect Unauthorized Access to Password ManagersInjection into or memory reads of KeePass, 1Password, LastPass or Bitwarden processes; vault files (.kdbx, .opvault) opened by other processesSecurity 4688; Sysmon 10 and 15; auditd; macOS unified log and osquery
T1555.006DET0130 Detect Unauthorized Access to Cloud Secrets Management StoresSecrets read by unusual identities, from unexpected regions, out of hours or at high volume, correlated with the preceding sign-inAWS CloudTrail GetSecretValue; equivalent Key Vault and Secret Manager audit logs

Making the endpoint analytics work

  • Turn on the right audit: Security 4663 for browser and vault files only appears if object access auditing is enabled and a SACL is set on those paths. Apply SACLs to the browser profile Login Data and Cookies files and Credential Manager folders on a pilot group first to measure event volume.
  • Allow-list the owner: the rule is "any process except the browser, its updater and approved backup or EDR agents". Unsigned binaries, scripting hosts and processes in temp or download folders reading these files are high-confidence.
  • Use the browser's own signal: Google states that Chrome App-Bound Encryption (Chrome 127 on Windows, cookies first, with passwords and other tokens planned) forces malware to gain system privileges or inject into Chrome, both more detectable, and that failed verifications log Event ID 257 from the Chrome source in the Application log.
  • Correlate with use: a store read followed by sign-ins from new infrastructure for accounts saved on that device is the strongest signal; feed infostealer log monitoring into the same case.

Making the cloud analytics work

  • Baseline per secret: most secrets are read by one or two workload roles. Alert on any new principal reading a secret, and on ListSecrets followed by many retrievals by one identity.
  • Keep the trail: the CloudTrail console lookup covers only 90 days, so deliver trails to central storage or a SIEM for longer hunts.
  • Pair with KMS: MITRE's IaaS analytic pairs GetSecretValue with KMS Decrypt; watch both.

Prevention and hardening

MitigationSub-techniquesConcrete action
M1027 Password PoliciesParent, .001, .003, .005Decide by policy whether browsers may store credentials; give the macOS Login Keychain a password different from the login password where risk justifies it; follow NIST guidance for master passwords
M1026 Privileged Account ManagementParent, .006Limit which accounts and services can query password stores and secrets managers, and to which secrets
M1042 Disable or Remove Feature or Program.004Enable "Network access: Do not allow storage of passwords and credentials for network authentication" where users do not need cached network credentials
M1054 Software Configuration.005Re-lock password managers after a short idle timeout so plaintext lives in memory briefly
M1021 Restrict Web-Based Content.003Restrict browser extensions and third-party cookies to shrink what a malicious extension or site can reach
M1051 Update SoftwareParent, .003, .005Keep browsers and password managers current to close extraction flaws
M1018 User Account Management / M1017 User Training.003, .005Remove inactive accounts; teach staff to keep corporate credentials out of personal profiles and to use the approved password manager

Settings that matter

  • Browser password saving: Chrome's PasswordManagerEnabled policy set to disabled stops users saving new passwords, but Google notes previously saved passwords still work, so pair it with a clean-up campaign. Control sign-in and sync to personal accounts on managed browsers. See the Browser Security Hardening Guide.
  • Windows network credentials: the policy above sits under Computer Configuration, Windows Settings, Security Settings, Local Policies, Security Options; Microsoft notes it needs a restart and may require users to type passwords for resources outside the domain.
  • Least privilege for local users: Google observes that App-Bound Encryption helps most where users cannot run downloaded files as Administrator, because that forces malware towards injection that endpoint agents catch.
  • Cloud secrets: scope read permissions to individual secrets or paths, not *; separate the right to manage a secret from the right to read its value; prefer workload identity federation and dynamic, short-lived credentials over stored static ones. See the Cloud Secrets Managers Guide.

Rollout order

  1. Measure: inventory how many managed devices have saved browser passwords, which domains they cover, and which cloud identities can read which secrets.
  2. Pilot detections: enable SACL-based file auditing and the browser, Credential Manager and vault analytics in alert-only mode on one device ring for two weeks; tune allow lists for EDR, backup and sync agents.
  3. Move credentials: migrate saved corporate logins to the enterprise password manager or SSO, then disable browser saving by policy on a pilot ring. Roll back by unsetting the policy; existing passwords are unaffected either way.
  4. Tighten cloud read rights: replace wildcard secret read grants with per-secret grants, using access logs to find real readers; keep a break-glass role with alerting.
  5. Automate response: on a confirmed store access, isolate the device, revoke sessions and rotate every credential saved on it, prioritising non-human ones.

Choosing where to start

If most staff use SSO and an enterprise password manager, disabling browser saving is low friction and removes the largest store. If not, deploy the password manager first, or you will push users towards spreadsheets. Small organisations get most value from browser policy, EDR and infostealer log monitoring; large ones should add SACL-based file auditing, secret-level cloud baselines and automated rotation.

Where identity fits

T1555 is where human device hygiene becomes a non-human identity problem. Developers and operators save cloud console passwords, API keys, database credentials and tokens in browsers and personal vaults because it is convenient. Treat any credential saved on an endpoint as exposed when that endpoint is compromised, move service credentials into a managed secrets platform with per-workload access, and give every secret an owner who can rotate it quickly. See the Secrets Management Guide.

Lessons from incidents

  • Old stealer logs stay dangerous: Mandiant reported that some Snowflake credentials had not been rotated for up to four years after the infection that exposed them.
  • Personal sync crosses the boundary: the Cisco case shows a corporate secret leaving managed control through a personal browser profile.
  • One vault, many keys: LastPass showed that a human vault holding cloud keys inherits the risk of that person's home device.

Framework mapping

FrameworkRelevant items
NIST SP 800-53 Rev. 5IA-5 Authenticator Management; AC-6 Least Privilege; CM-7 Least Functionality; SC-28 Protection of Information at Rest; SI-4 System Monitoring
NIST CSF 2.0PR.AA-01 identities and credentials managed; PR.PS-01 configuration management practices applied; DE.CM-09 computing hardware, software and their data monitored
ISO/IEC 27001:2022 Annex A5.17 Authentication information; 8.1 User endpoint devices; 8.2 Privileged access rights; 8.16 Monitoring activities
CIS Controls v8.1Control 4 Secure Configuration; Control 5 Account Management; Control 10 Malware Defenses
Related ATT&CKT1003.001 LSASS Memory; T1539 Steal Web Session Cookie; T1552.005 Cloud Instance Metadata API; T1078.004 Cloud Accounts; T1212 Exploitation for Credential Access

Practitioner checklist

  • Decide and enforce by policy whether managed browsers may save passwords and sync to personal accounts.
  • Provide an enterprise password manager with short auto-lock and SSO-based unlock.
  • Enable object access auditing on browser credential files and Credential Manager folders, and deploy the DET0037 and DET0134 analytics.
  • Monitor macOS for unexpected security and osascript use and keychain reads.
  • Disable storage of network credentials where not required.
  • Scope cloud secret read permissions per secret and alert on new readers and bulk retrieval.
  • Subscribe to infostealer log monitoring for corporate domains and act on hits within hours.
  • On any confirmed store access, rotate all credentials saved on the device, non-human ones first.

Standards and references

Related NHI Mgmt Group resources: Infostealer Threat Guide · ATT&CK Credential Access Tactic Guide · Cloud Secrets Managers Guide · Browser Security Hardening Guide

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 27 October 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.