Security

How TalentTwin protects accounts, candidate data and assessment results. This page describes controls built into the product today.

Accounts and authentication

Password storage
Passwords are hashed with bcrypt using a per-password salt. Plain-text passwords are never stored.
Session tokens
Sign-in issues short-lived access tokens and refresh tokens delivered in HttpOnly cookies, so page scripts cannot read them. Tokens are revoked on sign-out.
Rate limiting
Sign-in, account and public lookup endpoints are rate limited per user or IP address to slow down credential stuffing and scraping.

Access control

Role-based permissions
Candidates, recruiters, institution staff and administrators have separate roles. Each API action checks the caller's effective permissions, and role changes within an organisation are recorded in an authorization audit log.
Organisation isolation
Employer and institution data is scoped to the organisation that owns it. Database sessions carry the tenant context used by PostgreSQL row-level security.
Candidate consent
Employers see a candidate's profile through the platform; contact details on public credentials are masked, and recruiter contact requests are relayed rather than exposing the candidate's email.

Your data and the DPDP Act, 2023

Consent you can withdraw
Processing consent is recorded per account and can be granted or revoked at any time.
Export and correction
Candidates can export the personal data held about them and request corrections.
Erasure
An erasure request deletes the uploaded CV file from storage, clears the profile's search embedding, and removes the candidate's profile data.
Retention
Assessment recordings and proctoring data are purged on a schedule based on each organisation's configured audit retention period.
Data residency
The deployment is configured for the ap-south-1 (India) region. See the privacy policy for details.
Grievances
Complaints can be raised with the Grievance Redressal Officer.

Assessment integrity and AI

Sandboxed code execution
Code submitted in assessments runs in an isolated sandbox with static checks before execution.
Proctoring signals
During proctored sessions the browser records signals such as tab switches, face presence and background audio anomalies. These are attached to the session for human reviewers.
Human decisions
Scores and rankings are advisory. Hiring decisions are made by people, and candidates can ask for a human re-evaluation.
Audit trails
AI model calls, score overrides and candidate record changes are written to audit logs so results can be traced back to their inputs.

Application security

Security headers
API responses set a Content-Security-Policy, deny framing and disable MIME-type sniffing.
Request tracing
Every request receives a trace ID, which lets us investigate incidents without logging request bodies.

Certifications

We do not list third-party certifications on this page. If your organisation needs a security questionnaire completed, or you want to report a vulnerability, email security@qinoc.ai.