One of the most challenging practical aspects of providing end-to-end network security for legacy client-server protocols such as non-anonymous FTP (File Transfer Protocol) is convincing end users to actually use the secure alternatives, rather than abandoning them in favor of simpler, more familiar, or more fully featured insecure clients. A number of secure alternatives to the FTP protocol have been developed, but thus far have met with only limited success -- we feel this is primarily due to the fact that these solutions almost universally require the end user to learn a new, unfamiliar client interface or tweak complicated settings in order to make the security work. The average end user is interested in maintaining the security of their account, but is unwilling to invest a significant effort to setup a complicated system or the time to learn a whole new interface. SafeTP is a unique new FTP security system that strikes at the heart of this problem by providing completely transparent FTP security for users of Microsoft Windows. SafeTP operates by installing a transparent proxy in the Windows networking stack which detects outgoing FTP connections from any legacy (insecure) Windows FTP client, and silently secures them using modern cryptographic techniques (the server must also support SafeTP in order for a secure connection to be successfully established). SafeTP is 100% compatible with existing (insecure) FTP servers, and will operate in an insecure mode if the server does not yet support the SafeTP protocol. One key feature of the SafeTP client proxy is that it was designed to be completely transparent to the client FTP application. This way, users can reap the benefits of FTP security, while continuing to use their existing FTP software. Since its recent release on the internet, SafeTP has become extremely popular and is rapidly gaining acceptance in a diverse user community that includes numerous corporations, educational institutions and private users. In this paper, we describe the design of SafeTP and our experiences in implementing and maintaining this successful system. We discuss various challenges encountered in designing a fully transparent and interoperable security layer, and the solutions we implemented. We also describe various aspects of the hybrid public-key and shared-key cryptosystem used to provide confidentiality, integrity, and authenticity for FTP sessions.

We present an “adaptive multi-start” genetic algorithm for the Euclidean traveling salesman problem that uses a population of tours locally optimized by the Lin-Kernighan algorithm. An all-parent cross-breeding technique, chosen to exploit the structure of the search space, generates better locally optimized tours. Our work generalizes and improves upon the approach of Boese et al. [2]. Experiments show the algorithm is a vast improvement over simple “multi-start,” i.e., repeatedly applying Lin-Kernighan to many random initial tours. Both for random and several standard tsplib [5] instances, it is able to find nearly optimal (or optimal) tours for problems of several thousand cities in a few minutes on a Pentium Pro workstation. We find these results are competitive both in time and tour length with one of the most successful TSP algorithms, Iterated Lin-Kernighan.