, which employ trusted certificate authorities (CAs) to establish certificate validity chains.
In the last years, a number of security concerns regarding PKI usage have arisen: certificates
can be issued for entities in the Internet, regardless of its position in the CA hierarchy tree. This
means that successful attacks on CAs have the potential to generate valid certificates enabling
man-in-the-middle attacks. The possibility of malicious use of intermediate CAs to perform
targeted attacks through ad-hoc certificates cannot be neglected and are extremely difficult to
detect.
Current PKI infrastructure for TLS is prone to MITM attacks, and new mechanisms for
detection and avoidance of those attacks are needed. IETF and other standardization bodies
have launched several initiatives to enable the detection of “forged” certificates. Most of these
initiatives attempt to solve the existing problems by maintaining the current PKI model and
using certificate pinning, which associates certificates and servers on use. These techniques
have significant limitations, such as the need of a secure bootstrap procedure, or pinning
requiring some host-by-host basis.
This study proposes an evolution from pinning-in-the-host to pinning-in-the-net, by enabling
mechanisms to validate certificates as they travel through a given network. Certificates would
be classifi ed as trusted or not trusted as a result of cross-information obtained from different
sources. This would result in early detection of suspicious certifi cates and would trigger
mechanisms to defeat the attack; minimize its impact; and gather information on the attackers.
Additionally, a more detailed and thorough analysis could be performed.