Abstract
Most critical security vulnerabilities depend on the OS. If a hacker finds a machine with a vulnerable OS, then he can attack the system. Network administrators can defend against OS-specific attacks if they can find vulnerable machines before hackers do, but physically checking or actively scanning a large network can take time and resources. This paper describes a modification of p0f implemented in the Linux kernel, called k-p0f, which is a tool for this problem. This paper describes the design of k-p0f and compares its performance to p0f with both laboratory-generated and real-world traffic.