1. Scope
In scope:
- the predictra.io website;
- the prod.predictra.io client application;
- the api2.predictra.io API.
Out of scope:
- our clients’ assets (domains, IP addresses, applications): report them to the client concerned, not to us;
- our providers’ services (cloud, payment, email, etc.): use their own disclosure programs;
- denial of service, load testing and deliberate flooding;
- social engineering (phishing, calls) targeting our staff or our clients;
- physical attacks and access to our premises;
- spam and automatically generated reports without proof of exploitability;
- simple deviations from best practices with no demonstrated impact (missing headers without an attack scenario, displayed versions, SPF or DMARC alone): we read them, but they fall outside this policy.
2. Rules of conduct
To benefit from our commitment (section 4), you must:
- test only with accounts you own or that we provided to you; never access another client’s data, nor modify, delete or keep it;
- stop immediately and notify us if you access personal information, leaked credentials, analysis results or any other client data; do not download, copy or disclose them;
- not use the platform to launch analyses against third parties, for example by bypassing domain verification;
- use the least intrusive proof of concept possible; do not install a backdoor or maintain access;
- keep a reasonable request rate;
- not disclose the vulnerability publicly before it is fixed or before the end of the time frame set in section 5;
- comply with applicable laws, including the Criminal Code (s. 342.1 and 430(1.1)), as well as those of the country you are testing from.
3. How to report
Write to contact@predictra.io. Our security.txt file (/.well-known/security.txt) points to the same address.
A good report contains: a description of the vulnerability, the addresses or API endpoints concerned, the steps to reproduce, the estimated impact, the date of discovery and a way to reach you (a pseudonym is fine).
4. Our commitment to good-faith researchers
If you follow this policy, Predictra:
- considers your research authorized, for the scope in section 1;
- will not bring civil action against you and will not file a complaint with the authorities for this research;
- will not hold it against you that you bypassed a technical measure strictly to the extent needed for the demonstration;
- will keep you informed of progress on the fix;
- will thank you publicly if you wish.
This commitment binds only Predictra. It binds neither our clients, nor our providers, nor public prosecutors, nor the authorities of another country. It does not cover activities that are out of scope or contrary to section 2.
5. Time frames
Faster disclosure can be agreed by mutual consent.
6. If the vulnerability exposed data
If personal information may have been accessed without authorization, we handle the event as a confidentiality incident: entry in the register, assessment of the risk of serious injury, notice to the Commission d’accès à l’information and to the persons concerned where applicable, and notice to the clients concerned. See our privacy policy.
7. Third-party flaw discovered while analyzing a client (summary)
Analyzing a client’s attack surface can reveal a weakness that belongs to a third party: shared hosting provider, content delivery network (CDN), cloud provider, software vendor. In that case:
- we do not push the analysis beyond what is needed to qualify the finding, and no active analysis is run on an asset whose control the client has not confirmed;
- we inform the client first and recommend that they notify their provider themselves;
- with the client’s written consent, we may notify the third party through its official channel, naming the client or not;
- a product flaw with no CVE identifier is disclosed in a coordinated way to the vendor, directly or through the Canadian Centre for Cyber Security or CERT/CC, without naming the client;
- we never name a client to a third party without its written consent, unless required by law.
8. Ransomware claims and third-party data (summary)
- A post on a leak site is a claim, not a fact. We present it as such until the targeted organization has confirmed it.
- We do not download, buy or keep stolen data: only the metadata of the claim (group, dates, victim name and domain, address of the post).
- We do not interact with malicious actors: no negotiation, no purchase.
- Credentials found in leaks or infostealer logs are communicated only to the organization whose domain is concerned, to protect its accounts. They are never used to sign in or tested.
- We do not directly contact the people whose information appears in this data. The client organization assesses its own obligations.