Anonymize log files for AI
To anonymize log files for AI before pasting them in is the habit that best protects technical teams. A log excerpt, a support ticket, a directory export or a configuration file contains IP addresses, user IDs, hostnames and sometimes an API key forgotten in a comment. Once redacted, that text lets you ask an online AI to explain an error while keeping the map of your information system to yourself. Biffer processes these files on the administrator's computer, with no connection, and replaces each identifier with a stable alias.
This page describes the files concerned, the applicable obligations, the AI uses and the fields of the IT profile.
Logs, tickets, exports: an IT department's files
Technical files have one particular feature: they mix ordinary personal data with machine identifiers that, put end to end, describe the infrastructure.
- Logs and traces (text, CSV): source and destination IP addresses, user IDs, hostnames, timestamps, error messages.
- Support tickets (CSV or PDF export): ticket number, requester, email, phone number, description of the workstation and the incident.
- Directory and inventory exports (Excel, CSV): accounts, employee IDs, MAC addresses, license numbers, serial numbers.
- Configuration files and scripts (text): IP addresses, hostnames, sometimes a hard-coded API key or token.
Biffer processes text (.txt, .md), CSV, Excel, Word and PDF files while preserving their structure: a log stays line by line, a CSV export keeps its columns. The PDF case is detailed on the page redact a PDF.
IP addresses, accounts, secrets: the IT department's obligations
An IP address, a user ID or a MAC address tied to a workstation is personal data under the GDPR as soon as it relates to an identifiable person: European data protection authorities hold this position consistently. Article 32 of the GDPR requires the controller and its processors to apply appropriate security measures, which for an IT department means controlled logging, access rights management and the protection of secrets.
Beyond the GDPR, three issues are specific to IT.
- Secrets. An API key, a token or a password pasted into an online AI must be considered compromised and revoked. Biffer spots them before anything is sent.
- Mapping. Hostnames, internal IP addresses and version numbers describe the infrastructure. Taken one by one, they seem harmless; gathered in a log, they make an attack easier.
- Contractual commitments. A managed service provider is bound by its clients' confidentiality clauses.
With offline software, everything stays on the administrator's computer. Biffer prepares, you validate: the administrator reviews the preview and the department has the final word on what is shared.
AI use cases and the Biffer workflow
Once pseudonymized, a technical file keeps all its value for an AI.
- Explain a stack trace taken from an application log.
- Summarize a hundred tickets exported as CSV by probable cause.
- Write an operating procedure from raw notes.
- Review a configuration file to find an inconsistency.
The workflow is the same every time.
- You drop the text, CSV, Excel, Word or PDF file into Biffer with the IT profile enabled.
- The local analysis highlights IP addresses, MAC addresses, user IDs, keys and tokens, ticket and license numbers, hostnames, as well as people, emails and phone numbers.
- You untick false positives (a loopback address, a product name) and add what is missing, all occurrences at once.
- The file comes out with stable aliases (IP_001, MACHINE_003, UTILISATEUR_002 for a user; the alias labels are in French): the log stays consistent, and the same machine keeps the same alias from one line to the next. In Word and PDF, redacted data is highlighted in black, an option enabled by default.
- The mapping table, encrypted or in plain text, restores the real values in the AI's answer so you act on the right machine.
The queue processes a series of logs in automatic mode. Biffer is free for five files per month, after a one-time email confirmation in the software; beyond that, Biffer for Mac or Biffer for PC costs €10 excl. VAT (€12 incl. VAT) per computer, paid once: unlimited files, lifetime license tied to the computer, all redaction features included. The steps are detailed on the features page.
The fields the IT profile recognizes
The IT profile adds the technical identifiers of the information system to the general detections. Here is the exact list, as it appears in the software.
| Field | Where you find it |
|---|---|
| IP address | Logs, configurations, audit reports, tickets |
| MAC address | Inventories, DHCP logs, network leases |
| Ticket number | Support tool exports, follow-up emails |
| User ID | Authentication logs, directory exports, access rights |
| API key / token | Configuration files, scripts, call traces |
| License number | Software inventories, vendor contracts, invoices |
| Hostname | Logs, diagrams, procedures, inventories |
On top of these come the four fields shared by all professions (file number, reference, contract number, employee ID) and the general detections: people, emails, phone numbers, postal addresses, French company numbers (SIREN and SIRET), French social security number, IBAN, license plates, dates. For an identifier specific to your organization, the custom keyword field described on the professions page does the rest. Biffer is published by Romain Rissoan, trainer, consultant and founder of OPCADIA, in Lyon.
Frequently asked questions
Are private and loopback IP addresses handled too?
Biffer detects any well-formed IPv4 and IPv6 address, including private ranges (10.x, 172.16 to 172.31, 192.168.x) and loopback. In an internal log, it is precisely the private addresses that describe the infrastructure, and they deserve to be replaced. If an address can stay visible, for example 127.0.0.1, you untick it in the preview and all its occurrences are kept. The alias assigned is stable: the same address gets the same IP_00n on every line, which lets the AI follow exchanges between two machines without knowing their real addresses.
Is an API key detected by Biffer still usable?
Yes, Biffer always preserves your original file: it produces a pseudonymized copy, and the key stays intact in the source. The detection serves as a warning. A key or token found in a file you were about to send to an AI has probably already circulated elsewhere: in a code repository, a ticket, an attachment. Good practice is to revoke and regenerate it, then store it in a secrets manager. Biffer relies on common key and token formats; for a format specific to your tool, a custom keyword field ("token=", "api_key:") completes the detection.
Can Biffer handle a log of several hundred thousand lines?
Large text and CSV files are processed line by line on your computer, the only limit being your machine's capacity, since processing is entirely local. Analysis time depends on the power of the computer and the number of detections. For a very long log, we advise validating the detections on an excerpt in reviewed mode, then running the full series through the automatic queue: the same rules apply to every file and aliases stay consistent from one file to the next. The output file keeps the same number of lines in the same order.
Your business documents anonymized, ready to be published or handed to an AI tool.
Biffer runs on your computer without sending a single file. Free for five files a month, unlimited for €10 excl. VAT per computer.
100% offline · Free: 5 files a month · No file sent