Helmet armour

As web attacks become far more common place, specifications are quickly getting ratified to stop the common attack vectors from being used. W3C has set up a Security Working Group just for improving security against website users.

The technologies mentioned here are all belt and braces security measures also known as defence in depth they therefore should not be considered the only defence against attacks.

Browsers that do support the standards, will gain from the increased security however browsers that don't are still vulnerable to the attacks. Even when all browsers do support the technology it is probably always safe to assume that they might not work.

Pinning behaviour to the browser

The web itself was designed to be stateless and so since lots of stateful technology solutions have been added on top, some of the new forms of security are in this realm and are also called TOFU security or trust on first use; obviously all modern tech needs a hipster name. TOFU is actually an old concept that came from SSH, by acknowledging you know a site, you are entrusting it on your first use. Despite its usage in browsers is not as secure as SSH, it is however another way to prevent a server from being stolen by an attacker after it was first used.

Tofu

SSH requests the user to approve unknown servers credentials this isn't the case for the web, mostly I suspect because of the level of impact to user experience. It is far from perfect but it does help prevent some attacks of trust against the users trust in the servers identity.

DNS poisoning and DNS hijacking are two ways in which an attacker can abuse the users trust in the server.

The following three technologies help solve these issues, allowing a site to pin behaviour to the browser for future requests:

  • (HPKP) Public key pinning - giving browsers the ability to pin what a sites certificates will be should make it much harder for an attacker to game the trust in a servers identity. Preventing man in the middle attacks will become much simpler with this, as a client will not trust a response with a different certificate from the original response.

  • CSP pinning - this has just been released as a working draft by the W3C and so there is no intent even from the browsers to implement to my knowledge. However it is exciting that a browser would be able to know, that it can immediately block all requests to specified resources from the site for a certain time window. This adds further hardening to CSP2 which gives the strongest security for resources loading within a page and also has become a recommendation last month. Perhaps this directive might also lead to CSP directives being able to be loaded externally via some form of manifest file, the header could then just contain the integrity of the file; that way as directives become more complex the overhead doesn't increase exponentially.

  • (HSTS) HTTP strict transport security - This isn't new however it fits into the same category as the two above. This allows sites to always enforce that they are in HTTPS mode only. Subsequent requests to the site will load HTTPS.

Preventing clickjacking

UI security is worth mentioning as a new set of directives for securing the interface of an application from malicious behaviour. These directives are additions to the CSP policy which add the following protections:

  • Prevention of obstructing page elements and triggering the user to click on items that were not expected.
  • Allowing certain elements or behaviours through.
  • Preventing time based attacks.
  • Preventing elements from being placed near certain elements.
  • Adding in a DOM interface to expose suspected attacks to JavaScript allowing the application to alert the user.
  • Further extending violation reports provided by CSP.

Fixing insecure resources

Upgrade Insecure Requests follows on nicely from HSTS, which tells the browser to force all insecure content to be HTTPS. For example the author may not be in control of all the sites content and can essentially rewrite the code to have 'https' in URLs that are 'http'.

Preventing information leaking

Referrer policy Provides a way to prevent browsers from supplying a referer header to other websites. This prevents any information leaks by URLs by using this header, this reduces the risk of users exposing information from where they have came from. It also adds further protection against the risk of unique tokens being exposed; despite not ideal to be in the URL in the first place certain interfaces require it.

Ensuring our content is correct

(SRI) Sub Resource Integrity Is a way to ensure that content served to the browser is of a known content shape. Sites serving integrity attributes, prevent browsers from executing the sub resources unless they match the identity specified in the HTML. This prevents scripts from being overridden with malicious content, its primary application is content served over a CDN, however it doesn't harm making attackers lives harder.

Another use case is restricting the content served from third parties, most sites often load many widgets and scripts from other sources - All of which may not remain secure over time. The developers of the application may not have a choice over if the resource is loaded on the site, for example it could be a primary revenue stream for the site. However the developer would be able to test the current code served from the third party and serve the integrity parameter from the script being served. If the resource then gets compromised the users of the site wouldn't be at risk of the attack.

Headers still worth using

  • X-XSS-Protection - XSS filtering for common techniques used for injecting XSS. It was never standardised, however it is acknowledged by both Chrome and Internet Explorer.

  • X-Content-Type-Options - Used to stop browsers from sniffing the content of an asset and running it, unless it matches the content type sent from the server. So for example loading text/plain content will never be sniffed and loaded as JavaScript.

  • X-Frame-Options Is a way of preventing clickjacking, by informing browsers when the site is allowed to be loaded through framing.

    • This has been integrated within CSP2 frame ancestors however it is still worth providing this until full support is available for CSP2.
  • Access-Control-Allow-Origin - A way of opening up a site to allow cross origin requests, it is worth noting this here as it impacts security. Unless you need CORS functionality then don't return this header.