Summary of Network Security
This chapter covers a very large and important area of computer security: networks and distributed applications. As the world becomes more connected by networks, the significance of network security will certainly continue to grow. Security issues for networks are visible and important, but their analysis is similar to the analysis done for other aspects of security. That is, we ask questions about what we are protecting and why we are protecting it. In particular, we ask
- What are the assets?
- What are the threats?
- Who are the threat agents?
- What are the controls?
- What is the residual, uncontrolled risk?
Network assets include the network infrastructure, applications programs and, most importantly, data. Recall that threats are actions or situations that offer potential harm to or loss of confidentiality, integrity, or availability, in the form of interception (eavesdropping or passive wiretapping), modification (active wiretapping, falsification, and compromise of authenticity), and denial of service. In stand-alone computing, most agents have a strong motive for an attack. But in networks we see new threat agents; anyone can be a victim of essentially a random attack. The strongest network controls are solid authentication, access control, and encryption.
Networks usually employ many copies of the same or similar software, with a copy on each of several (or all) machines in the network. This similarity, combined with connectivity, means that any fault in one copy of a program can create vulnerabilities spread across many machines. Mass market software often has flaws, and each flaw can be studied and exploited by an attacker. In large networks, a huge number of potential attackers can probe the software extensively; the result is that a network often includes many identified faults and software patches to counter them.
In a sense, security in networks is the combination and culmination of everything we know about security, and certainly everything we have discussed in this book so far. A network's security depends on all the cryptographic tools at our disposal, good program development processes, operating system controls, trust and evaluation and assurance methods, and inference and aggregation controls.
Networks and their security remind us that good software engineering practices can go a long way toward making software difficult to attack. When a network and its components are structured, designed, and architected well, the resulting system presents solid defenses and avoids potential single points of failure. And a well-engineered network is easy to change as it evolves; because it is easier to understand, changes seldom introduce unintentional flaws.
Many of the controls useful for stand-alone systems are also useful in networks. But three controls are specific to networks: firewalls, intrusion detection systems, and secure e-mail. These controls have evolved from many years of research, both in security and in other computer science realms. They emphasize why we should know not only the history of security but also the relevance of other computing research. For example, firewalls are just an updated form of reference monitor. Similarly, intrusion detection profits from more fundamental research into pattern-matching and expert systems. And secure e-mail is really a carefully designed application of cryptography. You might think that controls such as these are the result of strokes of genius. But in fact, they reflect the long-term nature of knowledge and engineering practice; new ways to provide security build on a growing base of understanding and experience.
Until now we have stressed technical controls, which can be very effective in protecting our computing assets. But many security losses come from trusted insiderseither honest people making honest, human mistakes or dishonest insiders able to capitalize on their knowledge or privileges. In the next chapter we consider administrative controls, such as security policies, user awareness, and risk analysis, as a way to address the insider threat.