The Information Commissioner's Office (ICO) has finally fixed a security bug on its website - five years after it was first notified to the data privacy watchdog.
IT consultant Paul Moore first warned the ICO about a cross site scripting (XSS) problem on its website in 2009. The flaw meant it was possible to introduce arbitrary content under the control of hackers while presenting it as if it had originated from the ICO's website, opening up the door to tricks that might be used to lend false authenticity to all manner of potential scams.
Moore blogged about the issue and other related problems involving the ICO in 2012.
The ICO stands for promoting best practice in data security, so the allegation that it was failing to keep its own house in order was well worth investigating. El Reg asked two penetration testing organisations for an opinion on Moore's warning.
One of those firms, Pen Test Partners, was able to confirm non-critical but still troublesome security flaws with the ICO's website.
"The < and > are not being properly escaped/encoded by the application, so it allows reflective cross site scripting to execute," explained Ken Munro, a senior partner at Pen Test Partners.
"However, this isn’t a desperately serious issue. One has to follow a link to the site. I accept that a link to the ICO would be more likely to be clicked than any old URL, but it still requires user interaction," he added.
Munro concluded that the ICO website was vulnerable to a reflected XSS, a notable but less serious flaw than a persistent XSS bug.
Armed with these findings El Reg requested a comment from the ICO. In response, the data privacy watchdog admitted there was a minor problem with the search field on its website while playing down the significance of the flaw and quibbling with the timing of the issue. The ICO added that it has fixed the bug anyway, something that Moore confirmed even before we received the ICO's statement.
Like any responsible organisation we take a risk based approach to security, adopting controls and countermeasures proportionate to the impact of any compromise and the threats that we are exposed to. To ensure that this objective is met we carry out regular testing of our systems to make sure that the security of the information processed is maintained.
Having considered the details provided to us we have addressed a minor issue relating to a search field on our website. This issue would not have been present when the blog you refer to was written, because the search technology now used was not introduced until a later date. We are satisfied that the other issues identified in the blog cannot be replicated.
We provide a public facing website which contains no sensitive information. We are confident that the integrity of our website has been maintained and that no ICO data has been compromised.
Munro noted that other XSS problems had cropped up on the ICO's website in the past (as recorded by xssed.org here).
"The ICO web site has anti-XSS measures in many places, but not all. They’ve clearly missed some," Munro added.
XSS is an all-too-common class of web security vulnerability. For an XSS attack to be successful, the victim must be tricked into following a link through some form of phishing, meaning some user interaction is required.
Moore was in broad agreement with Munro's assessment, while disputing some of the detail.
"Ken [Munro]'s right in so much as reflected attacks require social engineering to work, but in terms of damage potential, there really is no difference," Moore explained. "Social engineering attacks are used to hijack domains, steal credentials, access servers (think WHMCS a few years ago) - there's obviously a tangible risk there."
"To make matters worse, the ICO have very recently fined BPAS £200,000 for failing to keep their site secure. In my opinion, they're in no position to comment on the security of other sites having been vulnerable since 2009," he added.
Moore remains highly critical of the ICO's general approach to web security.
"For an authority entrusted with advocating and enforcing data privacy, best practice methodology and openness, I'm astonished by their lack of mitigative action and nonchalant attitude," Moore said.
"The ICO site has changed considerably since 2009, that much is true. However, the exploit and their misunderstanding of the risk remains unchanged. If it's possible to remotely alter content shown under the genuine ICO domain, an attacker can leverage the trust associated with that domain. The site may not contain sensitive information, but it can easily be altered to collect it from unsuspecting users," he added. ®