Computer Security: Art and Science - Hardcover

HORTENSIO:Madam, before you touch the instrumentTo learn the order of my fingering,I must begin with rudiments of artTo teach you gamouth in a briefer sort,More pleasant, pithy and effectual,Than hath been taught by any of my trade;And there it is in writing, fairly drawn.-; The Taming of the Shrew, III, i, 62-68 On September 11, 2001, terrorists seized control of four airplanes. Three were flown into buildings, and a fourth crashed, with catastrophic loss of life. In the aftermath, the security and reliability of many aspects of society drew renewed scrutiny. One of these aspects was the widespread use of computers and their interconnecting networks. The issue is not new. In 1988, approximately 5,000 computers throughout the Internet were rendered unusable within 4 hours by a program called a worm. While the spread, and the effects, of this program alarmed computer scientists, most people were not worried because the worm did not affect their lives or their ability to do their jobs. In 1993, more users of computer systems were alerted to such dangers when a set of programs called sniffers were placed on many computers run by network service providers and recorded login names and passwords. After an attack on Tsutomu Shimomura's computer system, and the fascinating way Shimomura followed the attacker's trail, which led to his arrest, the public's interest and apprehension were finally aroused. Computers were now vulnerable. Their once reassuring protections were now viewed as flimsy. Several films explored these concerns. Movies such as War Gamesand Hackersprovided images of people who can, at will, wander throughout computers and networks, maliciously or frivolously corrupting or destroying information it may have taken millions of dollars to amass. Reality intruded on Hackers when the World Wide Web page set up by MGM/United Artists was quickly altered to present an irreverent commentary on the movie and to suggest that viewers see The Net instead. Paramount Pictures denied doing this. Another film, Sneakers, presented a picture of those who test the security of computer and other systems for their owners and for the government. Goals This book has three goals. The first is to show the importance of theory to practice and of practice to theory. All too often, practitioners regard theory as irrelevant and theoreticians think of practice as trivial. In reality, theory and practice are symbiotic. For example, the theory of covert channels, in which the goal is to limit the ability of processes to communicate through shared resources, provides a mechanism for evaluating the effectiveness of mechanisms that confine processes, such as sandboxes and firewalls. Similarly, business practices in the commercial world led to the development of several security policy models such as the Clark-Wilson model and the Chinese Wall model. These models in turn help the designers of security policies better understand and evaluate the mechanisms and procedures needed to secure their sites. The second goal is to emphasize that computer security and cryptography are different. Although cryptography is an essential component of computer security, it is by no means the only component. Cryptography provides a mechanism for performing specific functions, such as preventing unauthorized people from reading and altering messages on a network. However, unless developers understand the context in which they are using cryptography, and unless the assumptions underlying the protocol and the cryptographic mechanisms apply to the context, the cryptography may not add to the security of the system. The canonical example is the use of cryptography to secure communications between two low-security systems. If only trusted users can access the two systems, cryptography protects messages in transit. But if untrusted users can access either system through authorized accounts or, more likely, by breaking in , the cryptography is not sufficient to protect the messages. The attack-ers can read the messages at either endpoint. The third goal is to demonstrate that computer security is not just a science but also an art. It is an art because no system can be considered secure without an examination of how it is to be used. The definition of a "secure computer" necessitates a statement of requirements and an expression of those requirements in the form of authorized actions and authorized users. A computer engaged in work at a university may be considered "secure" for the purposes of the work done at the university. When moved to a military installation, that same system may not provide sufficient control to be deemed "secure" for the purposes of the work done at that installation. How will people, as well as other computers, interact with the computer system? How clear and restrictive an interface can a designer create without rendering the system unusable while trying to prevent unauthorized use or access to the data or resources on the system? Just as an artist paints his view of the world onto canvas, so does a designer of security features articulate his view of the world of human/machine interaction in the security policy and mechanisms of the system. Two designers may use entirely different designs to achieve the same creation, just as two artists may use different subjects to achieve the same concept. Computer security is also a science. Its theory is based on mathematical constructions, analyses, and proofs. Its systems are built in accordance with the accepted practices of engineering. It uses inductive and deductive reasoning to examine the security of systems from key axioms and to discover underlying principles. These scientific principles can then be applied to untraditional situations and new theories, policies, and mechanisms. Philosophy Key to understanding the problems that exist in computer security is a recognition that the problems are not new. They are old problems, dating from the beginning of computer security and, in fact, arising from parallel problems in the noncomputer world . But the locus has changed as the field of computing has changed. Before the mid-1980s, mainframe and mid-level computers dominated the market, and computer security problems and solutions were phrased in terms of securing files or processes on a single system. With the rise of networking and the Internet, the arena has changed. Workstations and servers, and the networking infrastructure that connects them, now dominate the market. Computer security problems and solutions now focus on a networked environment. However, if the workstations and servers, and the supporting network infrastructure, are viewed as a single system, the models, theories, and problem statements developed for systems before the mid-1980s apply equally well to current systems. As an example, consider the issue of assurance. In the early period, assurance arose in several ways: formal methods and proofs of correctness, validation of policy to requirements, and acquisition of data and programs from trusted sources, to name a few. Those providing assurance analyzed a single system, the code on it, and the sources vendors and users from which the code could be acquired to ensure that either the sources could be trusted or the programs could be confined adequately to do minimal damage. In the later period, the same basic principles and techniques apply, except that the scope of some has been greatly expanded from a single system and a small set of vendors to the worldwide Internet . The work on proof-carrying code, an exciting development in which the proof that a downloadable program module satisfies a stated policy is incorporated into the program itself, is an example of this expansion. It extends the notion of a proof of consistency with a stated policy. It advances the technology of the earlier period into the later period. But in order to understand it properly, one must understand the ideas underlying the concept of proof-carrying code, and these ideas lie in the earlier period. As another example, consider Saltzer and Schroeder's principles of secure design. Enunciated in 1975, they promote simplicity, confinement, and understanding. When security mechanisms grow too complex, attackers can evade or bypass them. Many programmers and vendors are learning this when attackers break into their systems and servers. The argument that the principles are old, and somehow outdated, rings hollow when the result of their violation is a nonsecure system. The work from the earlier period is sometimes cast in terms of systems that no longer exist and that differ in many ways from modern systems. This does not vitiate the ideas and concepts, which also underlie the work done today. Once these ideas and concepts are properly understood, applying them in a multiplicity of environ-ments becomes possible. Furthermore, the current mechanisms and technologies will become obsolete and of historical interest themselves as new forms of computing arise, but the underlying principles will live on, to underlie the next generation-;indeed the next era-;of computing. The philosophy of this book is that certain key concepts underlie all of computer security, and that the study of all parts of computer security enriches the understanding of all parts. Moreover, critical to an understanding of the applications of security-related technologies and methodologies is an understanding of the theory underlying those applications. Advances in the theory of computer protection have illuminated the foundations of security systems. Issues of abstract modeling, and modeling to meet specific environments, lead to systems designed to achieve a specific and rewarding goal. Theorems about composability of policies and the undecidability of the general security question have indicated the limits of what can be done. Much work and effort are continuing to extend the borders of those limits. Application of these results has improved the quality of the security of the systems being protected. However, the issue is how compatibly the assumptions of the model and theory conform to the environment to which the theory is applied. Although our knowledge of how to apply these abstractions is continually increasing, we still have difficulty correctly transposing the relevant information from a realistic setting to one in which analyses can then proceed. Such abstraction often eliminates vital information. The omitted data may pertain to security in nonobvious ways. Without this information, the analysis is flawed. The practitioner needs to know both the theoretical and practical aspects of the art and science of computer security. The theory demonstrates what is possible. The practical makes known what is feasible. The theoretician needs to understand the constraints under which these theories are used, how their results are translated into practical tools and methods, and how realistic are the assumptions underlying the theories. Computer Security: Art and Science tries to meet these needs. Unfortunately, no single work can cover all aspects of computer security, so this book focuses on those parts that are, in the author's opinion, most fundamental and most pervasive. The mechanisms exemplify the applications of these principles. Organization The organization of this book reflects its philosophy. It begins with mathematical fundamentals and principles that provide boundaries within which security can be modeled and analyzed effectively. The mathematics provides a framework for expressing and analyzing the requirements of the security of a system. These policies constrain what is allowed and what is not allowed. Mechanisms provide the ability to implement these policies. The degree to which the mechanisms correctly implement the policies, and indeed the degree to which the policies themselves meet the requirements of the organizations using the system, are questions of assurance. Exploiting failures in policy, in implementation, and in assurance comes next, as well as mechanisms for providing information on the attack. The book concludes with the applications of both theory and policy focused on realistic situations. This natural progression emphasizes the development and application of the principles existent in computer security. Part 1, "Introduction," describes what computer security is all about and explores the problems and challenges to be faced. It sets the context for the remainder of the book. Part 2, "Foundations," deals with basic questions such as how "security" can be clearly and functionally defined, whether or not it is realistic, and whether or not it is decidable. If it is decidable, under what conditions is it decidable, and if not, how must the definition be bounded in order to make it decidable? Part 3, "Policy," probes the relationship between policy and security. The definition of "security" depends on policy. In Part 3 we examine several types of policies, including the ever-present fundamental questions of trust, analysis of policies, and the use of policies to constrain operations and transitions. Part 4, "Implementation I: Cryptography," discusses cryptography and its role in security. It focuses on applications and discusses issues such as key management and escrow, key distribution, and how cryptosystems are used in networks. A quick study of authentication completes Part 4. Part 5, "Implementation II: Systems," considers how to implement the requirements imposed by policies using system-oriented techniques. Certain design principles are fundamental to effective security mechanisms. Policies define who can act and how they can act, and so identity is a critical aspect of implementation. Mechanisms implementing access control and flow control enforce various aspects of policies. Part 6, "Assurance," presents methodologies and technologies for ascertaining how well a system, or a product, meets its goals. After setting the background, to explain exactly what "assurance" is, the art of building systems to meet varying levels of assurance is discussed. Formal verification methods play a role. Part 6 shows how the progression of standards has enhanced our understanding of assurance techniques. Part 7, "Special Topics," discusses some miscellaneous aspects of computer security. Malicious logic thwarts many mechanisms. Despite our best efforts at high assurance, systems today are replete with vulnerabilities. Why? How can a system be analyzed to detect vulnerabilities? What models might help us improve the state of the art? Given these security holes, how can we detect attackers who exploit them? A discussion of auditing flows naturally into a discussion of intrusion detection-;a detection method for such attacks. Part 8, "Practicum," presents examples of how to apply the principles discussed throughout the book. It begins with networks and proceeds to systems, users, and programs. Each chapter states a desired policy and shows how to translate that policy into a set of mechanisms and procedures that support the policy. Part 8 tries to demonstrate that the material covered elsewhere can be, and should be, used in practice. Each chapter in this book ends with a summary, descriptions of some research issues, and some suggestions for further reading. The summary highlights the important ideas in the chapter. The research issues are current "hot topics" or are topics that may prove to be fertile ground for advancing the state of the art and science of computer security. Interested readers who wish to pursue the topics in any chapter in more depth can go to some of the suggested readings. They expand on the material in the chapter or present other interesting avenues. Roadmap This book is both a reference book and a textbook. Its audience is undergraduate and graduate students as well as practitioners. This section offers some suggestions on approaching the book. Dependencies Chapter 1 is fundamental...