Experts poke holes in crypto protocol for more secure web

Adoption of HTTPS has skyrocketed across the web over the past year, with the majority of global websites now making use of the encrypted communication protocol.

But some say that HTTPS has not been entirely effective at making the web truly secure, and that poor implementation of the HTTP security extension has produced various cryptographic flaws ready for attackers to exploit.

“In the near future, everyone is going to use HTTPS,” said Stefano Calzavara, an assistant professor in computer science at the Università Ca’ Foscari Venezia in Italy.

“So the next part of the process to make the web more secure would be banning unsafe HTTPS implementations.”

The complexity of the web, where the security of one site is reliant on the robustness of several different subdomains and linked services (be it an advertiser or web authentication method), has created an environment where vulnerabilities can become amplified.

“So even if a single part of HTTPS is deployed incorrectly, it might be that the security of the entire web application is compromised,” Calzavara explained.

While vulnerabilities within the HTTPS web ecosystem have been known for some time, the extent of their current impact remains unclear.

“We know these attacks work but it’s not obvious [exactly] where they still work,” Calzavara said.

Calzavara and his colleagues in Venice have now quantified the threat to the SSL/TLS protocol in a paper to be presented at the IEEE Symposium in San Francisco next week.

“One thing about the work was trying to understand what the implications of prescript cryptographic flaws on web application security are,” said Calzavara, speaking to The Daily Swig ahead of this year’s conference.

“We found that these problems are sometimes underestimated,” he added.

Calzavara and his team assessed three exploitable HTTPS vulnerabilities found throughout the top 10,000 websites, as rated by Alexa.

Nearly 900 of these sites were found to be vulnerable to TLS attacks, including the homepages of household names including Alibaba, Deutsche Bank, Verizon, and LinkedIn.

The HTTPS vulnerabilities analysed were selected based on their severity and popularity, Calzavara said – including exploits that dealt in page integrity, cookie-based authentication, and web tracking.

“Cookies are used for web authentication, websites load content from other sources, which is the page integrity issue, and a lot of services use advertisements,” he said.

One in 10 of searched websites had page integrity issues, which occurs when a script is sent to domain from a ‘tainted’ channel.

“So if I’m Google, and even if my HTTPS is correct, I may be loading a script from, say, Yahoo, and if Yahoo has not deployed HTTPS correctly, some attacker can compromise the integrity of the connection,” Calzavara said.

412 websites were exposed to possible cookie theft and session hijacking, while 543 websites were subject to cookie integrity attacks – issues that would allow a malicious actor to impersonate a user or further compromise the integrity of the website itself.

20% of websites were also found to be susceptible to having their cookies leaked, likely due to the practice of sharing cookies between different domains.

“If the implementation of HTTPS in the subservice is not implemented correctly, the attacker might be able to steal cookies [set] by Google because a vulnerability in the subservice might be able to reveal cookies to the attacker,” Calzavara said.

“This means an attacker is able access your calendar, your Gmail, [and] access whatever service Google provides.”

Poor HTTPS implementation of web trackers also proved to be prone to abuse, with 142 websites exposing users to profiling attacks.

Calzavara said that the main limitation to the work, however, was the fact that the team could not run the attacks, citing ethical reasons.