Security at SENDFU

Practical controls for accounts and API access.

This page describes security controls currently implemented for the SENDFU website, account system, API credentials, administrative access, and supporting operations.

Current implementationAccount and API controlsLast reviewed: August 20, 2026

Identity and credential controls

Customer credentials are separated from service-side credentials. Access to inference is reviewed and is not granted merely by registering an account.

Passwords

Salted password hashes

Account passwords are processed with a salted password-hashing function. SENDFU does not store the submitted plaintext password as the account credential.

API access

Hash-only API key storage

Issued API keys are stored as cryptographic hashes with non-secret prefixes and lifecycle metadata. The full key is presented when issued rather than retained for later display.

Sessions

Signed browser sessions

Web sessions use signed cookies with Secure, HttpOnly, and SameSite attributes. Invalid signatures and sessions beyond the configured lifetime are rejected.

Request and access boundaries

Controls are applied at both the web-account boundary and the authenticated API boundary.

Web forms

CSRF protection

State-changing account forms require a signed anti-CSRF value. Request bodies are also subject to a size limit.

Authentication

Login rate limiting

Repeated login attempts are rate-limited, with account-level failed-attempt handling and temporary lock controls.

Authorization

Role-separated administration

Administrative account review, API-key state changes, and usage exports are separated from ordinary account access and require an authorized administrator role.

API-key and service controls

SENDFU-issued inbound credentials can be constrained independently of internal service credentials.

Key state
Only active keys are accepted. Keys can be suspended or revoked.
Expiry
A configured expiry is enforced; invalid or expired values fail closed.
Model scope
A key may be limited to an explicit model allowlist, and non-allowlisted model IDs are rejected before service processing.
Quota and accounting
Accepted requests are checked against configured usage limits and recorded for operational usage accounting when required usage counters are available.
Credential separation
Clients authenticate with SENDFU-issued keys. Service-side credentials remain server-side and are not accepted from client requests.

Operational safeguards

The current service includes bounded controls intended to reduce common web and operational risks. These controls do not eliminate all risk.

Browser boundary

Security headers

Public responses use controls including Content Security Policy, frame restrictions, content-type sniffing protection, and a restrictive referrer policy.

Data integrity

Backup and recovery controls

The account store uses atomic replacement and restrictive file permissions. Backup and restore tooling validates the stored structure, and production startup fails closed on a damaged store.

Visibility

Status monitoring

The public status system checks service and model-catalog availability and retains sampled history. Catalog checks are not represented as successful real-time inference tests.

Assurance boundary

Security statements should be read as descriptions of the current implementation, not as a substitute for contract-specific review.

Current assurance position.

No internet service can guarantee absolute security. SENDFU does not claim SOC 2 or ISO 27001 certification on this page, and does not claim a third-party penetration-test program, PCI DSS scope, 24-hour security operations center, or guaranteed incident-response time unless separately documented and evidenced.

Responsible reporting

Report suspected credential exposure, unauthorized account activity, or a reproducible security issue through the current support channel.

Security-sensitive report

Contact SENDFU

Email business@sendfuglobal.com with a concise description, affected URL or component, reproduction steps, and any non-sensitive evidence.

Do not include passwords, API keys, inference content, or other secrets in the report.

Prepare security report