Mobile Bot Detection With
Less Data Exposure
Cynopta's minimized off-device data flow is not a feature - it is a core design constraint. Cynopta decides its score on the phone: none of the inputs it scores has to reach Cynopta's servers. What the SDK sends to Cynopta afterwards depends on your settings and is listed below and in our data-flow documentation.
Scored on the Device
The score is decided entirely on the phone, with no network call and without sending its inputs to Cynopta. Afterwards the SDK sends aggregated statistics for licensing and billing and, unless you switch them off, dashboard reports with detailed scoring results and crash reports. Our data-flow documentation lists every field, recipient, purpose and retention period.
We Don't Pool Customers' Data
We use customers' data for training or research only with design partners whose agreement explicitly allows it, and we never mix different customers' data.
Designed to Run Without Permission Prompts
Cynopta is designed to operate without user-facing permission prompts, subject to the final platform execution and signal specification. Seamless for the end user: no extra prompts or friction. Customers keep control of their own notices.
Built for a Stricter Standard
Data-Minimizing Architecture
Cynopta is designed to minimize off-device data flows and to avoid advertising identifiers and cross-app tracking. That is a data-minimizing starting point under GDPR, CCPA, India's DPDP Act and Brazil's LGPD, not a settled compliance position: we are still finalizing our recommended defaults for apps with users in the EEA and the UK. Customers remain responsible for determining lawful basis, notices, permissions, and platform disclosures for their own deployment.
Designed Without Cross-App Tracking
Cynopta does not trigger Apple's App Tracking Transparency (ATT) framework and uses no advertising identifiers or cross-app tracking. The iOS SDK, in internal beta and not yet available to customers, carries an up-to-date privacy manifest and does not track users under Apple's definition. For Android, we give you the answers you need for Google Play's Data safety form. Your app's own store disclosures remain your responsibility.
Minimal Off-Device Footprint
Many cloud bot platforms build their consoles on raw IPs, per-request logs, and user-agent forensics - data your DPO must defend. With Cynopta, the score doesn't need raw behavioral data to leave the device, and every report the SDK sends is documented.
Privacy as Competitive Edge
Stronger fraud detection without increased data exposure, regulatory risk, or user trust erosion. Cynopta turns privacy compliance into a commercial advantage - particularly in regulated markets where cloud behavioral vendors face DPO objections and data transfer restrictions.
Touch, sensor and timing data are turned into a score on the device. None of it has to reach Cynopta's servers for the score to be decided.
Thousands of measurements are processed locally by on-device AI
After scoring, the SDK sends only the reports listed below: aggregated statistics for licensing and billing and, unless you switch them off, detailed results for your dashboard and crash reports. Each is described in our data-flow documentation.
Every Report, Listed
The score itself needs no network call. Separately, the SDK sends these reports to Cynopta, and on Android it asks Google for an integrity check. The full field list is in our data-flow documentation and in Part B of our Privacy Policy.
Daily assessment counts, score-band counts and weekly summaries for licensing and billing. It carries the license ID, which identifies your app's license, and a report ID that is random and new for each report. No ID linked to the device or its user.
Detailed scoring results and on-device summaries for your dashboard and analytics, with a hashed installation ID, the device model and the carrier. Three modes: full report (the default), security-related data only, or off.
Errors in the SDK's own code only, sent to Sentry in Europe, which does not store IP addresses.
The SDK gives Google a one-way hash of the usage report and gets back an integrity token, so Google is a recipient. We delete the token within two hours.
Raw screen taps, such as touch coordinates, used only for debugging and internal research. Release builds stop sending them when you choose security-related data only or switch dashboard reports off.
Names, email addresses, phone numbers, contacts, messages, advertising IDs, or the content of photos and videos. The SDK counts the photos taken each day, but never collects or analyzes them.
| Permission (install-time) | Why |
|---|---|
| INTERNET | Licensing and reports |
| ACCESS_NETWORK_STATE | Network context |
| ACCESS_WIFI_STATE | Network context |
| CHANGE_WIFI_MULTICAST_STATE | Nearby-device summaries |
| BLUETOOTH | Nearby-device summaries |
| RECEIVE_BOOT_COMPLETED | Resume after a restart |
None of these prompts the user
All six are standard install-time permissions. The SDK asks for no location, contacts, camera or microphone access.
Storage on the device
Data on the phone is kept on a rolling basis, from 14 to 365 days depending on its type, and deleted by the SDK's stop call. On Android, the main database is excluded from device backups.
Signals that can be switched off
The Android SDK has switches that turn individual signals off, such as installed apps, Wi-Fi, Bluetooth and the home network scan. Turning a signal off also removes the summaries that depend on it. All signals, including the home network scan and the work network route test, are on by default. Access to the switches through the SDK's API is being added.
Storage on our servers
Reports are stored with AWS, Cloudflare and Neon, all in North Virginia, United States, and crash reports with Sentry in Europe. We keep them for 13 months by default, unless you ask us for a longer or shorter period.
Questions to Ask Any Vendor
Ask every bot and fraud SDK you evaluate the same questions. Our answers are on this page and in our data-flow documentation.
- Does any raw behavioral data leave the phone to produce the score?
- What does the SDK send, to whom and how often, and which parts can we switch off?
- Which identifiers travel with the data, and do they survive a reinstall?
- Where is the data stored, and for how long?
- Is our data ever pooled with other customers' data or used to train shared models?
- Which permissions does the SDK declare, and does it ever prompt the user?