The Art of Computer Virus Research and Defense

nazarijo writes "I think by now we're all familiar with viruses and worms. It may have been a term paper diskette chewed up by a virus back in college, a family member's computer infected with the latest worm, or your email inbox clogged with a mass mailer of the week. But how do AV researchers dissect such malware, especially when virus writers have devoted so much time to avoiding detection and perfecting their craft with self-decrypting viruses, polymorphic shellcode, and obfuscated loops. Haven't you wanted a peek into how that's done, and how you would analyze such a monster that landed in your computer? Well, Peter Szor's book The Art of Computer Virus Research and Defense (TAOCVRD) has been gaining lots of critical acclaim lately for filling that gap, and rightfully so. (Before we begin, however, I should make one thing perfectly clear: I was a technical reviewer of this book. I enjoyed it when I read it originally, and I'm even more pleased with the final result. And now on to your regularly scheduled review.)" Read on for the rest. The Art of Computer Virus Research and Defense author Peter Szor pages 713 publisher Addison Wesley Longman and Symantec Press rating 9 reviewer Jose Nazario ISBN 0321304543 summary Clear, sweeping coverage of virus history and technical details

TAOCVRD opens with Part 1: Strategies of the attacker. Here we get to start to think about malicious code from the original ideas and viewpoints of its makers. Chapter 1 opens up with various games of the classic computer science world, including Conway's Game of Life and Core Wars, which is still fun after all of these years. From this we can start to think about computer viruses as a natural extension of other self-replicating computer structures. What's great about this chapter is that you can actually understand, and share in, the fascination of replicating code. It's as if you can understand the pure world that some virus writers live in.

Chapter 2 starts off the virus-analysis section, including some of the basics (like the types of malicious programs and their key features), as well as the naming scheme. Chapter 3, "Malicious Code Environments," serves as a lengthy and complete description of how various viruses work. The dependencies that you would expect to see, including OS, CPU, file formats, and filesystems, are all described. Then Szor goes on to describe how viruses work with various languages, from REXX and DCL to Python and even Office macros. Not all of the descriptions are lengthy, but you get to see how flexible the world of writing a virus can be. What I most enjoyed about the book overall is represented in this chapter, namely Szor's command of the history of the virus as well as his technical prowess, which he drops in as appropriate.

Chapter 4 gets a bit more technical and now focuses on infection strategies. Again, Szor isn't afraid to delve into history or technical meat, including a lengthy and valuable section "An In-Depth Look at Win32 Viruses." If you don't feel armed to start dissecting viruses by this point, you're in luck: there's so much more to read. Chapter 5 covers in-memory strategies used by viruses to locate files, processes, and sometimes evade detection. Szor has a list of interrupts and their utility to the virus writer, providing a comprehensive resource to the virus analyst.

Chapters 6 and 7 cover basic and advanced self protection schemes, respectively, used by viruses. TAOCVRD's completeness of information in a usable space, together with very functional examples and descriptions, is again evident. Szor walks you through a basic decryptor routine, for example, showing you how a self-contained virus can be both evasive and functional at the same time. Sadly little attention is given to various virus construction kits at the end of chapter 7, though.

Chapters 8 and 9 get a little less technical and somewhat more historical. These chapters cover virus payloads and their classification (ie benevolent viruses, destructive viruses, etc) and computer worms, respectively. The overview of payloads is almost entirely historical, giving a great overview of how virus writers have used their techniques to cause havoc or just have "fun" from time to time. Chapter 9 gives a concise and valuable overview of computer worms, almost boiling about half of my worms book down into just one chapter in a clear and easy to use fashion.

Part 1 concludes with chapter 10, which covers exploits and attack techniques used by worms and viruses. Again, Szor's clarity of explanation shines as he artfully gives a concise overview of how a buffer overflow attack works (including stack layout and address manipulation), heap-based attacks, format string attacks, and related methods. He then discusses these techniques in light of various historical examples, clearly explaining how they operated and were successful. If you've been yearning for a short overview of attack techniques and how malware has used them, this chapter is for you.

Part 2 covers the defender's strategies. Chapter 11 serves as a nice introduction to this section by describing many of the current and advanced defense techniques such as some of the first and second generation scanners, code and system emulation, and metamorphic virus detection. This is all covered in nice technical detail, always at a reasonable level to not leave everyone in the dust. Through it all small examples are constantly given, which reinforce the text nicely. Chapter 12 is very similar, this time focusing on in-memory scanning and analysis techniques.

Chapter 13 covers worm blocking techniques, focusing on host-based methods which can prevent the buffer overflow from being successful or the code from arbitrarily gaining network access again. Chapter 14 complements this with network specific defenses, including ACLs and firewalls, IDS systems, honeypots, and even counterattacks. These two chapters are a lot less technical than the previous two, but still quite valuable.

By this point I'm sure you're ready to try your hand at virus analysis, and Szor is eager to help you out. In chapter 15 he gives you a great setup for virus analysis, including various tools and examples of how they work and what kind of information they give you. Finally, in chapter 16 you have the obligatory (and valuable) resource roundup which complements the references given in every chapter, as well.

Overall I find Szor's book to be amazing, both in terms of its technical prowess over so many specifics in the field but also for its presentation. Without dumbing it down, Szor's able to communicate to most readers with clarity in a manner they'll understand, learn from, and be able to use. I think that many of us, especially those of us who get plundered in our email inboxes with malware, are curious to spend some time dissecting these beasts using techniques AV professionals use, and Szor's book does an exemplary job of introducing that world to us all. I consider this to be one of the most important computer security books I own due to it's clarity and completeness of coverage.

You can purchase The Art of Computer Virus Research and Defense from bn.com. Slashdot welcomes readers' book reviews -- to see your own review here, read the book review guidelines, then visit the submission page.

Comments

Practice makes perfect

[kaosrain]

But how do AV researchers dissect such malware, especially when virus writers have devoted so much time to avoiding detection and perfecting their craft with self-decrypting viruses, polymorphic shellcode, and obfuscated loops.

It is important to note that virus authors who have perfected their craft create viruses that are not found.

Too Much Joy

[Gary W. Longsine]

I think you have hit the nail on the head here.

Reverse engineering malware is so much fun, and appeals to techie and tech-savvy manager types so much that it has been a terrific and terrible distraction. I've seen the effect firsthand -- companies waste precious limited mitigation and response talent and time trying to analyze malware when they should be taking immediate action to contain the spread of a worm.

Corporations and government agencies have been so thoroughly trained by the AntiVirus industry that they have a hard time coping in an age of the zero day worm, flash worm, or even the boring ordinary retread worm with 800 variants that do different things and propagate through a dozen different old defects. In fact, in the last year it's become clear that worms targeting many old defects can spread widely, slipping in under the radar of AntiVirus definitions with dozens of daily variants. (It's hard to patch a large network, and the industry hasn't woke up to the fact that it's also hard to keep it patched.)

What does it matter, which of the 800 strains of Spybot or Rxbot is smacking your PC's around? Well, if it were possible to quickly assess exactly what a given strain might do on a computer, it might be. But typically it's not possible.

In fact, it's gotten to the point where the AntiVirus vendors themselves have all but given up on detailed analysis of the many variants emerging each hour. Sometimes critical features of a strain (what ports does it probe, etc.) are missing entirely from the public analysis of the strain for weeks after it was first detected. Sometimes one vendor will describe a feature while others don't. Obvious cut-and-paste errors in the analysis of major vendors can also be observed, if one pays close attention.

The AntiVirus industry can't keep up the analysis of every minor strain, but they do continue the practice because it's a proven effective strategy for keeping mindshare. To their credit, they do a pretty reasonable job of rapid analysis and signature development on quite a few variants every day. Unfortunately, the stakes are pretty high and getting higher.

The bottom line for big networks: focus on prevention and containment. Cleanup is very costly, so do your own analysis if you must, but don't let it delay or sap resources from containment efforts when a worm hits. Other damages might be mounting while the mitigation effort stalls out because an incident response team is bogged down trying to answer the question: "Does the variant that hit your network today have a keystroke logger?"

With several variants of various worms released each day, are you *sure* that you've been hit with only one variant?

Even if you think you are sure, in fact, you typically can't be sure quickly enough. Well staffed, well funded, and highly experienced labs at the major AntiVirus vendors can't keep up with detailed analysis of the zillions of variants. Neither can the overburdened IT staffs of the world. They need to stop trying.

Disclaimer: As the founder of Intrinsic Security I am clearly convinced enough in the limitations of the AntiVirus approach that I started a company and developed an alternative (complementary) approach. All of my opinions, well reasoned and otherwise, are my own, although they may be shared by others.