Local compromise

Local compromise means an attacker already has code execution, file access, or equivalent control on the endpoint where the browser runs. Once that boundary is crossed, browser-only protections against remote web attacks are often insufficient because the attacker can target stored secrets and trusted process state.

Expanded Definition

Local compromise describes a state, not a single exploit: the attacker has already gained execution, read access, or equivalent control on the endpoint that hosts the browser or related application state. At that point, browser boundaries no longer provide strong protection against theft of session material, tokens, cached data, or other secrets that were assumed to be protected by remote-only attack models. The term is especially important in identity and AI-adjacent environments because local compromise can expose credentials, API keys, and authenticated browser sessions that enable lateral movement into SaaS, admin consoles, or agent tooling. For that reason, it sits closer to endpoint trust loss than to classic website vulnerability categories. NIST’s Cybersecurity Framework is useful here because it treats endpoint protection, detection, and recovery as connected governance outcomes rather than isolated browser settings. Definitions vary across vendors when local compromise is discussed in product documentation, so NHI Management Group uses it strictly to mean the endpoint itself has already been breached. The most common misapplication is treating local compromise as a browser feature problem, which occurs when teams assume client-side hardening alone can defend secrets after hostile code is already running.

Examples and Use Cases

Implementing controls for local compromise rigorously often introduces friction for users and operations, requiring organisations to weigh stronger containment against convenience, persistence, and workflow continuity.

• A phishing payload installs a remote access trojan on a developer laptop, allowing the attacker to read browser session storage and reuse authenticated access to internal SaaS tools.
• A stolen device with unlocked profile access lets an intruder extract cached tokens, certificate material, or password manager data before the endpoint is remotely wiped.
• Malicious browser extensions or injected process memory capture active web sessions even when the user has strong remote-authentication controls in place.
• In an agentic AI environment, endpoint malware may steal API keys or local tool credentials used by an AI agent, turning a compromised workstation into a launch point for automated misuse.
• Incident responders may reference Anthropic’s report on AI-orchestrated cyber espionage when assessing how local endpoint control can be combined with automation to scale abuse.

In practice, local compromise often appears after the initial intrusion is already complete, when investigators discover that the attacker did not need to break the browser itself because the endpoint had become a trusted execution surface.

Why It Matters for Security Teams

Security teams need to understand local compromise because many browser and SaaS protections assume the endpoint is trustworthy enough to preserve secret state. Once that assumption fails, session hijacking, token replay, clipboard theft, memory scraping, and secret extraction become realistic next steps. This is particularly relevant for identity teams managing NHI, since browser-based admin portals, cloud consoles, and agent control planes often rely on stored credentials or short-lived tokens that are valuable only until the endpoint is lost. Guidance from the OWASP Top 10 remains helpful for remote web attack paths, but local compromise pushes the problem into endpoint hardening, token minimisation, and recovery planning. NIST CSF is relevant because it links protection, detection, and incident response around compromised assets rather than assuming the browser can preserve trust on its own. Organisations typically encounter the operational impact only after a session is abused or an endpoint is seized, at which point local compromise becomes operationally unavoidable to address.

Related resources from NHI Mgmt Group

• What breaks when a supply chain compromise can read local developer secrets?
• How do I respond to a confirmed NHI credential compromise?
• How do attackers turn a supply-chain incident into wider NHI compromise?
• What did Shai Hulud 2.0 actually compromise?