Here's a rough outline of the sort of stuff to consider around nonfunctional security requirements.

Foremost is identifying and classifying who is using the system in the first place from a functional perspective. These actors will likely fall out directly from the use cases and other requirements modelling exercises so you'll have end users, business users, administrators, editors, whatever... These will probably turn into roles in the security model of the system. ok, cool.

So assuming everyone stays nice and friendly, never lies, and obeys the rules you lay down then you can probably go home now - job done! Erm, but they don't and you need to assume that every actor; human or system, is a liar and a cheat, wanting to steal your data, corrupt your system, bend the rules in their favour and so on. So a short list as a starter for ten of topics which need to be considered - let me know what more needs to be added:

  1. Authentication and authorisation - How user identities are authenticated and authorised - who are they and what can they do?

  2. Identity and access management - How identities and access control is managed. From password reset policies to identity providers.

  3. Auditing and logging - The recording of events for analysis and accountability; including for non-repudiation purposes.

  4. Encryption - What do you encrypt and how do you do it? Much more than just SSL - which is a good start over the public network - you've got to worry about the insiders as well and; fearing the worst, what if your defences are breached and data is stolen? Using weak cyphers and home-grown algorithms is nearly as bad as storing sensitive data in plain text.

  5. Security zones and network zones - Creating distinct zones to protect critical resources, limiting access accordingly and auditing every gnats fart that occurs in your data-centre. Employing network appliances such as firewalls, reverse proxies and network intrusion detection systems accordingly.

  6. Secure development practices - From preventing code injection attacks to avoiding buffer overflow problems and employing defensive programming techniques. The Open Web Application Security Project (OWASP) has excellent resources on everything relating to software security.

  7. Denial of Service - How to avoid and deal with DDoS attacks (Distributed Denial of Service attack)? In the 21st century public services and businesses are utterly dependent on the net and cyberterrorism will become increasingly prevalent as the first frontier of attack.

No comments:

Post a Comment

Voyaging dwarves riding phantom eagles

It's been said before... the only two difficult things in computing are naming things and cache invalidation... or naming things and som...