Antivirus software makes Greg Shipley so mad he has to laugh. "The relationship between signature-based antivirus companies and the virus writers is almost comical. One releases something and then the other reacts, and they go back and forth. It's a silly little arms race that has no end."
Shipley, chief technology officer at Neohapsis, a security consultancy in Chicago, says the worst part is that the arms race isn't helpful either to him or his clients. "I want to get off of signature-based antivirus as rapidly as possible. I think it's a broken model, and I think it's an incredible CPU hog."
The question is, where should he go? Antivirus as an industry has modeled itself on the human immune system, which slaps a label on things like viruses so it knows to attack them when it sees that same label, or signature, again. Signature-based antivirus has moved well beyond that simple type of signature usage (though at the beginning, it did look for specific lines of code). In its current, more sophisticated form, it dominates the market for security software, despite some obvious limitations: You don't use it to stop data leakage, for instance, though many kinds of malware are designed to siphon data out of companies. The number of malware signatures tracked by security software company F-Secure doubled in 2007, and while you might cynically expect such a company to say there's more malware out there, 2007's total doubled the number of signatures F-Secure had built up over the previous 20 years.
Even before 2007, there were plenty of people besides Shipley arguing that antivirus was an industry in trouble. In fact, in 2006, Robin Bloor, an analyst at Hurwitz & Associates, penned a report titled "Anti-virus is dead." He argued that malware exists only because antivirus software exists, and said that antivirus software was doomed to be replaced by new forms of software, which he calls application control, or software authentication tools. Such tools whitelist the software we use and won't run anything else without the user's explicit permission.
Antivirus firms think their death is greatly exaggerated, thank you very much -- even those that aren't overly reliant on signatures, like BitDefender, which says that signature-based techniques account for only 20% of the malware it catches.
"Signatures aren't dead -- you need them," says Bogdan Dumitru, CTO at the Romanian firm, which uses behavioral targeting techniques to stop the remainder of attacks. Its main research focus is to develop an "undo" feature that will let users hit by malware reverse its effects. BitDefender hopes to release this feature in 2008.
Meanwhile, Bit9, the application whitelisting company highlighted in Bloor's report, uses antivirus software to help build its database -- 22 kinds of antivirus software, in fact. In November 2007, it announced a deal to give access to this database to security software maker Kaspersky Labs. Bit9 officials said that the database will help Kaspersky check new signatures to limit false positives.
It's also true that antivirus makers continue to sell billions of dollars worth of software, despite Bloor's proclamation. Bloor, though, says that "the technique of protecting PCs using virus signatures is now on the wane," and he rattles off a list of whitelisting companies offering software authentication tools -- not just Bit9, but also companies such as Lumension (formerly SecureWave), Savant Protection, Computer Associates and AppSense. And he noted the Kaspersky deal and Apple's use of whitelisting to protect the iPhone.
Not just whitelisting
Antivirus software has its uses. If a system is actually infected by malware, it "may be the least painful way of removing it," says David Harley, administrator of Avien, the antivirus information exchange network, adding, "Whitelisting does seem to be advocated currently as the panacea du jour. I think this relentless search for The Answer, discarding one partially successful solution set for something else in the hope that it will eliminate the problem, is actually unprofessional."
Harley makes that argument because he doubts that any single technology approach will be a 100% solution when it comes to security. He wrote that whitelisting thus is likely a supplemental technology for fighting malware, making it one of a host of newer technologies that have been adopted, including heuristics, sandboxing and behavior monitoring.
Corporate CISOs certainly don't expect to find one answer to their problems. "If you rely on signatures for security, you're pretty much dead in the water," says Ken Pfeil, head of information security for the Americas region of WestLB, a German bank. Pfeil thinks signatures are useful and his firm uses them. But when new malware appears, he often finds it faster to try to break it down himself to understand its potential effects, rather than wait for his vendor to give him an update. His firm has also adopted tools that use heuristics techniques and anomaly testing, to add oomph to its antivirus approach.
That kind of layered approach to software fits with where Natalie Lambert, an analyst at Forrester Research, thinks the market is going. She says that signature-based antivirus is "table stakes" for security software, and techniques like heuristic information processing systems, or HIPS, which look for suspicious actions by software, like an application opening itself from the Temp folder.
Lambert says McAfee is probably furthest along in using HIPS among the big antivirus makers, having had more time than its rivals to new features added via corporate acquisitions.
The downside to these technologies is that none are as simple and alluring as the old signature-based antivirus, which she called a "set it and forget it" technology. She notes that HIPS technologies are difficult to manage and will never be as simple as the old model, though she expects they will get easier over time.
Neohapsis's Shipley says none of these techniques are really new -- he notes that it's been more than four years since McAfee purchased Entercept, for instance. But "what role does it play, and what percentage of things does it stop? I have no visibility into that." Shipley says he plans to bring in Bit9 to look at whether it could really replace his current antivirus software.
Antivirus firms agree that they are becoming something different.
Sophos, for instance, uses several additions to signature-based AV. Sophos examines program behavior -- the modifications a program makes to things like system configuration and files as the program runs. The company has also built in a pre-execution algorithm, a kind of crystal ball to simulate what unfamiliar code looks likely to do.
Richard Wang, manager of Sophos Labs in the U.S., says that while signatures are easy to create, things like pre-execution code are harder and thus take more time. But the payoff is that it can work against multiple strains of malicious software. He said that for the Storm worm, Sophos generated only one signature but has been able to recognize all the variants. Wang describes this type of technique as "almost like a broad-spectrum antibiotic."
Child's play?
Interestingly, the OLPC XO (from the One Laptop Per Child Foundation) is another place to look at new AV techniques. The XO uses the Bitfrost specification, developed expressly for this simple computer. OLPC claims that the system "is both drastically more secure and provides drastically more usable security than any mainstream system currently on the market."
The OLPC XO ships in a default mode that is basically locked down but simple for the user to open up. The Bitfrost specification uses a series of built-in protections, including sandboxes or program jails for applications and system-level protections that prevent alterations from code that could do something harmful.
Whether Bitfrost would work in a corporate environment or will be commercialized outside the OLPC project is unclear. But Avien's Harley, for one, thinks that there are psychological reasons why antivirus software is unlikely to go away.
"The idea of a solution that stops real threats and doesn't hamper nonmalicious objects and processes is very attractive. People (at any rate, those who aren't security specialists) like the idea of threat-specific software, as long it catches all incoming malware and doesn't generate any false positives, because then they can just install it and forget about it. Unfortunately, that's an unattainable ideal."
Note to Greg Shipley: Don't hold your breath on getting rid of your antivirus software.
Shipley, chief technology officer at Neohapsis, a security consultancy in Chicago, says the worst part is that the arms race isn't helpful either to him or his clients. "I want to get off of signature-based antivirus as rapidly as possible. I think it's a broken model, and I think it's an incredible CPU hog."
The question is, where should he go? Antivirus as an industry has modeled itself on the human immune system, which slaps a label on things like viruses so it knows to attack them when it sees that same label, or signature, again. Signature-based antivirus has moved well beyond that simple type of signature usage (though at the beginning, it did look for specific lines of code). In its current, more sophisticated form, it dominates the market for security software, despite some obvious limitations: You don't use it to stop data leakage, for instance, though many kinds of malware are designed to siphon data out of companies. The number of malware signatures tracked by security software company F-Secure doubled in 2007, and while you might cynically expect such a company to say there's more malware out there, 2007's total doubled the number of signatures F-Secure had built up over the previous 20 years.
Even before 2007, there were plenty of people besides Shipley arguing that antivirus was an industry in trouble. In fact, in 2006, Robin Bloor, an analyst at Hurwitz & Associates, penned a report titled "Anti-virus is dead." He argued that malware exists only because antivirus software exists, and said that antivirus software was doomed to be replaced by new forms of software, which he calls application control, or software authentication tools. Such tools whitelist the software we use and won't run anything else without the user's explicit permission.
Antivirus firms think their death is greatly exaggerated, thank you very much -- even those that aren't overly reliant on signatures, like BitDefender, which says that signature-based techniques account for only 20% of the malware it catches.
"Signatures aren't dead -- you need them," says Bogdan Dumitru, CTO at the Romanian firm, which uses behavioral targeting techniques to stop the remainder of attacks. Its main research focus is to develop an "undo" feature that will let users hit by malware reverse its effects. BitDefender hopes to release this feature in 2008.
Meanwhile, Bit9, the application whitelisting company highlighted in Bloor's report, uses antivirus software to help build its database -- 22 kinds of antivirus software, in fact. In November 2007, it announced a deal to give access to this database to security software maker Kaspersky Labs. Bit9 officials said that the database will help Kaspersky check new signatures to limit false positives.
It's also true that antivirus makers continue to sell billions of dollars worth of software, despite Bloor's proclamation. Bloor, though, says that "the technique of protecting PCs using virus signatures is now on the wane," and he rattles off a list of whitelisting companies offering software authentication tools -- not just Bit9, but also companies such as Lumension (formerly SecureWave), Savant Protection, Computer Associates and AppSense. And he noted the Kaspersky deal and Apple's use of whitelisting to protect the iPhone.
Not just whitelisting
Antivirus software has its uses. If a system is actually infected by malware, it "may be the least painful way of removing it," says David Harley, administrator of Avien, the antivirus information exchange network, adding, "Whitelisting does seem to be advocated currently as the panacea du jour. I think this relentless search for The Answer, discarding one partially successful solution set for something else in the hope that it will eliminate the problem, is actually unprofessional."
Harley makes that argument because he doubts that any single technology approach will be a 100% solution when it comes to security. He wrote that whitelisting thus is likely a supplemental technology for fighting malware, making it one of a host of newer technologies that have been adopted, including heuristics, sandboxing and behavior monitoring.
Corporate CISOs certainly don't expect to find one answer to their problems. "If you rely on signatures for security, you're pretty much dead in the water," says Ken Pfeil, head of information security for the Americas region of WestLB, a German bank. Pfeil thinks signatures are useful and his firm uses them. But when new malware appears, he often finds it faster to try to break it down himself to understand its potential effects, rather than wait for his vendor to give him an update. His firm has also adopted tools that use heuristics techniques and anomaly testing, to add oomph to its antivirus approach.
That kind of layered approach to software fits with where Natalie Lambert, an analyst at Forrester Research, thinks the market is going. She says that signature-based antivirus is "table stakes" for security software, and techniques like heuristic information processing systems, or HIPS, which look for suspicious actions by software, like an application opening itself from the Temp folder.
Lambert says McAfee is probably furthest along in using HIPS among the big antivirus makers, having had more time than its rivals to new features added via corporate acquisitions.
The downside to these technologies is that none are as simple and alluring as the old signature-based antivirus, which she called a "set it and forget it" technology. She notes that HIPS technologies are difficult to manage and will never be as simple as the old model, though she expects they will get easier over time.
Neohapsis's Shipley says none of these techniques are really new -- he notes that it's been more than four years since McAfee purchased Entercept, for instance. But "what role does it play, and what percentage of things does it stop? I have no visibility into that." Shipley says he plans to bring in Bit9 to look at whether it could really replace his current antivirus software.
Antivirus firms agree that they are becoming something different.
Sophos, for instance, uses several additions to signature-based AV. Sophos examines program behavior -- the modifications a program makes to things like system configuration and files as the program runs. The company has also built in a pre-execution algorithm, a kind of crystal ball to simulate what unfamiliar code looks likely to do.
Richard Wang, manager of Sophos Labs in the U.S., says that while signatures are easy to create, things like pre-execution code are harder and thus take more time. But the payoff is that it can work against multiple strains of malicious software. He said that for the Storm worm, Sophos generated only one signature but has been able to recognize all the variants. Wang describes this type of technique as "almost like a broad-spectrum antibiotic."
Child's play?
Interestingly, the OLPC XO (from the One Laptop Per Child Foundation) is another place to look at new AV techniques. The XO uses the Bitfrost specification, developed expressly for this simple computer. OLPC claims that the system "is both drastically more secure and provides drastically more usable security than any mainstream system currently on the market."
The OLPC XO ships in a default mode that is basically locked down but simple for the user to open up. The Bitfrost specification uses a series of built-in protections, including sandboxes or program jails for applications and system-level protections that prevent alterations from code that could do something harmful.
Whether Bitfrost would work in a corporate environment or will be commercialized outside the OLPC project is unclear. But Avien's Harley, for one, thinks that there are psychological reasons why antivirus software is unlikely to go away.
"The idea of a solution that stops real threats and doesn't hamper nonmalicious objects and processes is very attractive. People (at any rate, those who aren't security specialists) like the idea of threat-specific software, as long it catches all incoming malware and doesn't generate any false positives, because then they can just install it and forget about it. Unfortunately, that's an unattainable ideal."
Note to Greg Shipley: Don't hold your breath on getting rid of your antivirus software.