Security and trust

How we protect your data and put safeguards around our analyses

A cybersecurity company should show its own practices. Here is where your data is hosted, how we safeguard analyses, protect accounts and data, and check the numbers we present.

Illustrative scene: a managed IT services team working calmly and with focus.
Illustrative scene

Hosting and subprocessors

Platform data is hosted in Canada. Some subprocessors, including the text generation service, process data outside Canada; their categories and processing locations appear in the privacy policy. Jira and GLPI integrations are chosen and configured by you.

Where is data processed?

Some subprocessors, including the text generation service, process data outside Canada, including in the United States. Their categories and processing locations appear in the privacy policy.

Which integrations do you choose?

Integrations you choose

ServiceData sentResponsibility
JiraTitle, severity, host, source and finding textChosen and configured by you; your access credentials are encrypted at Predictra
GLPISame fieldsYour instance, configured by you

Analyses with safeguards

Before any active analysis, you prove control of the domain. Vulnerability testing requires a signed authorization letter. The free scan is a passive review of the domain.

How do you prove domain control?

Before any active analysis, you prove that you control the domain through a DNS TXT record, which covers subdomains, a file published on the website, or an email to an administrative address for the domain. The proof is checked again before every analysis.

When is an authorization letter required?

Vulnerability testing requires an authorization letter signed by the asset owner. Intrusive testing also requires a dedicated option, a time window and an emergency contact. The letter is signed electronically or uploaded as a signed copy, then reviewed by our team.

How is the organization verified?

Every organization is verified by our team before it is opened: legal identity, the signatory’s authority and a sanctions check.

Who can stop a test?

The owner or an administrator of the organization, or of its managed service provider (MSP), can stop a vulnerability test or an application test in progress at any time.

What mandate is required for an MSP’s clients?

An MSP analyzes a client only after uploading its signed mandate, verifying one of the client’s domains and accepting the client itself.

What is outside the scope of our analyses?

The scope excludes brute-force and dictionary password attacks, sign-in with a leaked credential, social engineering, persistent access, deliberate denial of service and data extraction beyond what is needed to confirm a vulnerability.

Domain control

The proof is checked again before every analysis.

Signed authorization

Vulnerability testing requires a letter signed by the asset owner.

Stop at any time

The owner or an administrator can stop a test in progress.

Seen “Predictra Scanner” in your logs?

This is an external attack surface analysis authorized in writing by the domain holder under a client mandate. You can ask us to verify it.

What does “Predictra Scanner” mean?

Our analyses identify themselves with the user agent “Predictra Scanner (+https://predictra.io/securite)”. If you see it in your server logs, it is an external attack surface analysis authorized in writing by the domain holder under a client mandate.

How can I verify an analysis?

Write to contact@predictra.io with the IP address and timestamp you observed. We confirm the analysis or stop it.

What do we collect?

Only what the authorized analysis requires.

Access and accounts

Defined roles, two-factor authentication and an access log safeguard accounts. Predictra staff access uses a stated, logged session; selected administrative operations take place outside a session.

How does two-factor authentication work?

App-based two-factor authentication is offered to all users and required at every sign-in once enabled. It is mandatory for all Predictra staff. Two-factor authentication uses an authenticator app; single sign-on (SSO) is not offered.

What roles and permissions are available?

Owner, Administrator, Technical (analysis and operations) and Executive (summary views only: executive view, action plan, reports and trends). Each user’s permissions are checked on every request; access that has not been granted is refused by default.

How are invitations and sign-ins managed?

Team members are invited by email and choose their own password; an administrator does not set another person’s password. Passwords are never stored in plain text. Sessions can be revoked: deactivating an organization or account ends sessions in progress. Repeated sign-in attempts are slowed down.

Can Predictra staff view your data?

To view an organization’s data, Predictra staff open an access session with a stated reason and time limit, recorded in your access log; administrative operations such as approval, compliance review and false-positive review take place outside a session.

What does the access log contain?

An access log, available to the owner and administrators, records views and exports of sensitive data, access sessions and Predictra team actions in your organization, changes to roles and two-factor authentication, decisions on false positives and analysis authorizations. It is protected against modification and deletion.

Owner and Administrator

Access log available for review.

Technical

Analysis and operations.

Executive

Summary views only: executive view, action plan, reports and trends.

Data protection

Leaked passwords are not retained. Secrets and verification documents are encrypted; each organization’s data is isolated.

What happens to leaked passwords?

Passwords found in leaks are not retained. As soon as they are collected, they are replaced with attributes useful for remediation—length, character types and resemblance to a default password—and with a non-reversible fingerprint specific to your organization, which only shows whether the same password comes up again.

Which data is encrypted and isolated?

Sign-in credentials for your tools (Jira, GLPI), verification documents and two-factor authentication secrets are encrypted. Each organization’s data is isolated. Searches in public ransomware claims are not specific to your organization; records provided by the leak database are tied to the domain, not to the organization.

How does retention work?

Retention periods are applied automatically by data category: leak records, hidden web content, closed findings and executive views.

What does a deletion request remove?

An owner can request deletion of their organization or an account. Deletion also erases conversations with the assistant. Supporting documents—letters, proofs and acceptances—and the access log are retained for legal reasons.

Can you export your data?

With the Professional plan, you can export all of your organization’s data as a JSON and CSV archive, available through a link with a limited lifetime. No leaked password is included.

Executive view and assistant

The “How it is calculated” panel shows where each amount comes from. Numbers in the executive view are inserted from the data and checked. This check covers the numbers and references, not the reasoning.

The proof: how is it calculated?

Every number traces back to an observation from the analysis, a cited public source or a declared parameter, shown with its calculation. The “How it is calculated” panel shows where each amount comes from. In the executive view, numbers are inserted from the analysis data, and an automated check rejects any sentence whose number, host, CVE or date does not match the cited facts. This check covers numbers and references, not the reasoning. If the text generation service is unavailable, no new executive view is produced and the previous one remains displayed.

What is sent to produce the executive view?

A file of facts is sent to a US-based subprocessor: the organization’s name, domains, hosts, technologies, CVEs, excerpts of finding descriptions, ransomware claims that target you—including group, victim name and source—risk and remediation estimates, hosts drawn from stolen credentials and counts of leaked credentials, with passwords excluded.

In the platform
Executive view using sample data: a summary generated from analysis facts.
Sample dataPredictra model — public sources

The assistant only reads

The assistant answers your questions about your Predictra data. It only reads, with exactly your permissions, and is unavailable to the Executive role.

What permissions and precautions apply to the assistant?

The conversational assistant answers your questions about your Predictra data. It only reads, with exactly your permissions, and is unavailable to the Executive role. Fields containing leaked credentials are removed before anything is sent to the subprocessor that produces the answers. Do not enter a secret—password or key—in your questions: it would be sent to that subprocessor.

Responsible disclosure

Found a vulnerability in Predictra? Our policy gives the address to use, the scope, the rules and our commitment to good-faith researchers.

How do I report a vulnerability in Predictra?

Found a vulnerability in Predictra? Our responsible disclosure policy gives the address to use, the scope, the rules to follow and our commitment to good-faith researchers. The security.txt file is also available.

Start with a free scan.

A passive review of the domain, based on public sources.