Exaforce Details Token Theft, ClickFix and Vishing Attacks—and Response Steps

· Original article ↗

Summary

Exaforce describes token theft, ClickFix lures and help-desk vishing cases, then recommends behavior-based detection and prebuilt, low-friction revocation actions to limit data theft after account compromise.

Key points

  • Exaforce says one stolen GitHub OAuth token pulled 76 repositories in about 15 seconds; its platform detected the activity within five minutes, but data was exfiltrated from 472 repositories before action.
  • The article says poisoned npm packages and infostealers can steal developer tokens, while GitHub may not alert on token misuse.
  • A reported ClickFix lure used a shared Claude conversation to deliver a command that installed a macOS stealer targeting the keychain, browser data and cloud credentials.
  • In reported vishing cases, attackers impersonated IT, obtained approval of real MFA prompts and used stolen sessions to download data; one case involved 140,000 files and a ransom demand.
  • The article recommends behavior-based detection that correlates identity-provider sign-ins with a user's history, including unfamiliar devices, networks and locations.
  • It advises automating lower-impact actions such as revoking sessions and removing newly registered MFA methods, while requiring human approval for higher-impact actions.
  • Revoking an identity-provider session may not terminate sessions or tokens already issued by connected services; the article provides API-based revocation examples for several platforms.

Article Details

Defense Focus
Detect rapid token misuse, credential theft, and session hijacking by correlating identity, endpoint, network, and application behavior, then contain confirmed incidents through appropriately authorized automated response.
Detection Methods
  • Correlate sign-ins with the user's historical networks, cities, devices, and identity-provider activity; flag novel combinations such as first-time locations, Linux devices, or previously unseen home ISPs.
  • Detect unusual GitHub token activity by combining new network or city, rapid repository archive downloads, and a user agent inconsistent with the account's normal operating system.
  • Monitor for newly registered MFA methods during suspicious sessions and correlate subsequent sign-ins or token use across identity providers and applications.
  • Use endpoint detection to identify user-executed install commands that launch credential-stealing activity, including collection of keychains, browser cookies, saved passwords, and cloud credentials.
  • Investigate tokens and credentials minted by a user to assess the impact of suspected infostealer activity.
  • Hunt for related activity across employees because multiple targets may be attacked using the same servers and domains.
Data Sources
  • GitHub token and repository activity, including archive downloads, user agents, and source networks
  • Identity-provider sign-in records and user history from Okta and Microsoft
  • MFA method registration and authentication events
  • Endpoint telemetry for command execution and credential or keychain access
  • Application session and token activity across SaaS, cloud, code, and data services
Platforms
  • Exaforce platform
Defensive Actions
  • Revoke compromised sessions and application-issued OAuth tokens; do not assume that revoking an identity-provider session invalidates sessions or tokens issued by connected applications.
  • Remove MFA methods registered during the flagged session and deactivate credentials that can be re-enabled.
  • Prepare and test application-specific revocation procedures and ensure the accounts executing them have the required permissions.
  • Automate low-friction containment actions for high-confidence findings; require human approval for password resets, account disablement, permanent deletion, or changes affecting production services.
  • Prioritize revocation coverage for services holding critical data, including code repositories, cloud environments, productivity platforms, AI services, and data platforms.
  • After suspected credential theft, assess all tokens and credentials minted by the affected user and revoke those at risk.

Indicators of compromise

TypeIndicatorContext
DOMAINcontoso[.]add-passkey[.]comPasskey-themed look-alike sign-in domain cited as an example of attacker-directed phishing.
DOMAINoskeysetup[.]comDomain hosting a custom vishing-related phishing kit described in the article.

MITRE ATT&CK

Threat Actors

Malware

Vendors

Products

Related Articles