Purdue CERIAS Researchers Find Vulnerability in Google Protocol
[This is posted on behalf of the three students listed below. This is yet another example of bad results when speed takes precedence over doing things safely. Good work by the students! --spaf]
As a part of an INSuRE project at Purdue University, PhD Information Security student Robert Morton and seniors in Computer Science Austin Klasa and< Daniel Sokoler conducted an observational study on Google’s QUIC protocol (Quick UDP Internet Connections, pronounced quick). The team found that QUIC leaked the length of the password potentially allowing eavesdroppers to bypass authentication in popular services like Google Mail or G-mail. The team named the vulnerability Ring-Road and is currently trying to quantify the potential damage.
During the initial stages of the research, the Purdue team found that the Internet has been transformed over the last five years with a new suite of performance improving communication protocols such as SPDY, HTTP/2 and QUIC. These new protocols are being rapidly adopted to increase the speed and performance of applications on the Internet. More than 10% of the top 1 Million websites are already using some of these technologies, including many of the 10 highest traffic sites.
While these new protocols have improved speed, the Purdue team focused on determining if any major security issues arose from using QUIC. The team was astonished to find that Google's QUIC protocol leaks the exact length of sensitive information when transmitted over the Internet. This could allow an eavesdropper to learn the exact length of someone's password when signing into a website. In part, this negates the purpose of the underlying encryption, which is designed to keep data confidential -- including its length.
In practice, the Purdue team found QUIC leaks the exact length of passwords into commonly used services such as Google's E-mail or G-mail. The Purdue team than created a proof-of concept exploit to demonstrate the potential damage:
Step 1 - The team sniffed a target network to identify the password length from QUIC.
Step 2 - The team optimized a password dictionary to the identified password length.
Step 3 - The team automated an online attack to bypass authentication into G-mail.
The Purdue team believes the root cause of this problem came when Google decided to use a particular encryption method in QUIC: the Advanced Encryption Standard Galois/Counter Mode (AES-GCM). AES-GCM is a mode of encryption often adopted for its speed and performance. By default, AES-GCM cipher text is the same length as the original plaintext. For short communications such as passwords, exposing the length can be damaging when combined with other contextual clues to bypass authentication, and therein lies the problem.Conclusion
In summary, there seems to be an inherent trade-off between speed and security. As new protocols emerge on the Internet, these new technologies should be thoroughly tested for security vulnerabilities in a real-world environment. Google has been informed of this vulnerability and is currently working to identify a patch to protect their users. As Google works to create a fix, we recommend users and system administrators to disable QUIC in Chrome and their servers by visiting this link. We also recommend -- independent of this issue -- that users consider enabling two step verification with their G-mail accounts, for added protection, as described here. The Purdue team will be presenting their talk and proof-of-concept exploit against G-mail at the upcoming CERIAS Symposium on 18 April 2017.Additional Information
This research is a part of the Information Security Research and Education (INSuRE) project. The project was under the direction of Dr. Melissa Dark and Dr. John Springer and assisted by technical directors a part of the Information Assurance Directorate of the National Security Agency.
INSuRE is a partnership between successful and mature Centers of Academic Excellence in Information Assurance Research (CAE-R) and the National Security Agency (NSA), the Department of Homeland Security and other federal and state agencies and laboratories to design, develop and test a cybersecurity research network. INSuRE is a self-organizing, cooperative, multi-disciplinary, multi-institutional, and multi-level collaborative research project that can includes both unclassified and classified research problems in cybersecurity.
This work was funded under NSF grant award No. 1344369. Robert Morton, the PhD Information Security student, is supported under the Scholarship For Service (SFS) Fellowship NSF grant award No. 1027493.Disclaimers
Any opinions, findings, or conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation, CERIAS, Purdue University, or the National Security Agency.