Alok Menghrajani

Previously: security engineer at Square, co-author of HackLang, put the 's' in https at Facebook. Maker of CTFs.

This blog does not use any tracking cookies and does not serve any ads. Enjoy your anonymity; I have no idea who you are, where you came from, and where you are headed to. Let's dream of an Internet from times past.

Home | Contact me | Github | RSS feed | Consulting services | Tools & games

Web browsers have come a long way in terms of security. For example, you can protect against cross site scripting attacks by enable a header called CSP (Content-Security-Policy). You can prevent a lot of CSRF attacks by checking the Origin header. Etc.

One area which continues to fail open is privacy. Most web application frameworks require implementing explicit privacy checks. The privacy check is there to see if a given user is allowed to view or edit a specific piece of content. Currently, if such a check is missing, the failure mode is to allow anyone to view or edit the data.

As a result, most web applications are filled with privacy bugs! From a security point of view, we call this kind of failure mode "fail open", and it's a bad thing.

Here are some ideas that might be worth exploring:

  • Some kind of proxy/middleware that sits between the web application and the database. The proxy could track who created every row in the database and enforce a unix-like (user/group/other) permission system).
  • A key-value or graph oriented datastore which enforces privacy checks and can be extended with custom checks.
  • A SQL database which implements the same type system as the web application. It could then disallow joining tables on unrelated columns.
  • Something leveraging Postgres' row level security.
  • etc.