15 ans de Zero Trust chez Google, de l’opération Aurora à BeyondProd

Zero Trust Implementation at Google: From Operation Aurora to BeyondProd

On Wed, Jan 13, 2010, Google publicly disclosed a major cyberattack known as Operation Aurora. This attack, attributed to Chinese hackers, targeted Google as well as other technology companies, NGOs, and journalists. The breach highlighted the limitations of traditional perimeter-based cybersecurity strategies. This event initiated a significant shift in Google's security strategy towards a Zero Trust approach.

Since 2010, Google has adopted a Zero Trust model, which operates on the principle of "never trust, always verify." This model involves systematically and contextually verifying every access attempt, without presuming any entity as trustworthy by default. Key elements of this strategy include minimizing privileges, early detection of compromises, and continuous verification.

Google has enhanced infrastructure security by integrating its proprietary Titan chip into custom-designed servers. Each server startup is signed, verified, and validated, ensuring that only machines that prove their integrity can communicate on the network. Additionally, Google has implemented the SALSA framework to manage software development pipelines, ensuring source code validation and deployment artifact verification.

Google has eliminated the use of VPNs through the BeyondCorp initiative, allowing employees to access internal resources securely from any location using strong authentication and multiple signals. The BeyondProd initiative extends these principles to production services and microservices, requiring identity verification for every component interaction, enforced by the ALTH protocol.

Google is applying the same Zero Trust principles to its AI models, ensuring security from creation through deployment. Each stage of the AI lifecycle requires verification, akin to traditional software.

Operational Guide: Deploying Zero Trust

Step Objective Recommended Tools

Identify risk areas and implicit access CMDB, Active Directory, vulnerability scanners

1. Existing Environment Mapping

Secure user identities MFA, FIDO2 (Yubikey, Titan), SSO (Okta, Azure AD)

1. Strong Authentication

Restrict access to the necessary minimum RBAC, ABAC, access audits

1. Least Privilege

Contextual access to internal resources Identity-Aware Proxy, Zscaler ZPA, Cloudflare Access

1. VPN-Free Access

Verify microservice identities mTLS, service mesh (Istio, Linkerd), certificates

1. Service Security

Ensure software supply chain trust Secured CI/CD, SALSA, SBOM, Binary Authorization

1. Code Integrity

Quickly identify compromises SIEM, SOAR, EDR, behavioral analysis

1. Detection and Response

Recommended Deployment : Incremental, based on functional priorities and identified risks.

Mantra to Remember : "Never trust, always verify."

D’après FrenchWeb.