PriAna is a privacy analysis tool based on various data science and artificial intelligence techniques that tries to identify privacy related data on your computer. It scans for various file types and supports various database systems including MS Access, MS SQL Server, MySQL, ODBC, Oracle, PostgreSQL and SQLite. It can be a valuable tool for GDPR assesment and audit.
PriAna uses various NLP techniques to find privacy related content (in the current version English and Croatian words are supported, but support for other languages is pending) from numerous filetypes including but not limited to csv, doc, docx, eml, epub, gif, jpg, jpeg, json, html, htm, mp3, msg, odt, ogg, pdf, png, pptx, ps, rtf, tiff, tif, txt, wav, xlsx, xls, mdb, rtf, and html.
In addition to filetypes it also scans for MIME types so renamed files are scanned as well. Besides textual formats it extracts textual information from images using OCR as well as speech from audio files using speech to text. Currently only on Linux it also detects human faces on images.
It identifies various archive types including images of virtual machines: ovf, ova, box, vmdk, vdi, vhd, img, raw, iso, zip, rar, tar, 7z. In the current version such archives have to be extracted and scanned manually.
PriAna searches for privacy related keywords (currently defined in priana.py), education related keywords and sexual orientation related keywords. It identifies emails, Croatian social security numbers (OIB and JMBG), IP addresses, MAC addresses, GPS coordinates and bank card numbers.
Using NLP it identifies names and locations, and using face detection it identifies images with human faces on them.
PriAna has been tested on Linux and Windows. In case you encounter problems on any of these platforms, please file an issue on GitHub, we will look into it as soon as possible. If you manage to run PriAna on any other platform, please let us know, we are happily accepting patches ;-)
PriAna outputs a report of all found files or tables in databases categorized into high, medium and low risk items. The output reports are stored in the data directory of PriAna.
PriAna also implements a small expert system for GDPR compliance currently only in Croatian that can help you asses all found data. This expert system is by no means a full audit and it does not guarantee in any way that you are GDPR compliant even if it states so!
In order to run PriAna from source you will need a working Python (>= 2.7) environment including a number of modules: SPADE (version 2), textract, gettext, ZODB, magic, nltk, SQLAlchemy, pyODBC, multiprocessing, psutil, tkinter, ttk, tkMessageBox, tkFileDialog, PIL, pickle, ctypes and face_recognition.
To run the recommendation expert system you will need a working installation of SWI Prolog
In order to use the database scanning features on Linux you will need to install UnixODBC, MDB-tables, MDB-export.
On Windows you will need VC++ Redistributable and Microsoft Access Redistributable Database Engine. Other dependencies include AntiWord, Oracle's instant client, pdftools, Tesseract-OCR and textract's node version, as well as a number of DLLs included in the source.
To build the binary a shell script is included. You will need pyinstaller in order to build a working executable. On Windows you will need to run it through CygWin and need to install AutoHotKey.
The 'build.sh' script accepts 3 arguments:
- Platform (linux or win)
- Destination directory where the bineries are to be stored
- Optional skip if the value is no then all is to be build else the compile of the expert system is skipped and the destination folder is deleted in full.
A pre-built binary is available for Windows: http://ai.foi.hr/priana-win.zip
As well as for Linux (Ubuntu): http://ai.foi.hr/priana-linux.tar.xz
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent