In March and April, small bunches of e-mail messages arrived at the offices of defense agencies and contractors in the U.S. and Europe. To recipients, the messages seemed credible: Each was addressed to a specific worker, with a valid return address within the organization and visual elements that made it look like internal e-mail. Too sparse and sophisticated to trip anti-spam filters, the messages exploited a previously unknown hole in Microsoft Word that allowed them to slip by anti-virus filters. Those recipients who were unlucky enough to open the e-mails’ malicious attachments unwittingly installed a Trojan horse, which used the Internet Explorer Web browser to report back, through the network firewall, to machines in China and Taiwan.
Phishing attacks such as this one are nothing new. Online scams that lure online banking and e-commerce customers to phony Web sites and trick them into giving up sensitive account information have been a mainstay of online criminals for years. However, the increase in so-called spear-phishing attacks is new, as is the increasing sophistication of the software they use to penetrate enterprise networks.
In the past year, the number of targeted attacks against companies has increased from one or two a week to one or more a day. Although those numbers might sound laughable compared with e-mail virus and spam campaigns, which can be measured in the millions of messages, spear-phishing attacks are much more dangerous, says Paul Wood, senior analyst at MessageLabs.
“These are not headed to the kind of addresses you harvest from the Internet,” Wood says. “These people have massive intelligence on organizations they want to penetrate. … The messages are specific to the organizations that they’re trying to get something from.”
That is usually intellectual property: software source code, design documents, or schematics. In the case of a defense contractor, however, the potential harm from lost intelligence outstrips the usual costs. What’s an enterprise to do?
With single-factor user names and passwords fast becoming an IT joke and traditional strong authentication products still expensive to buy and deploy, enterprises are looking for new ways to make authentication smarter, more pervasive, and easier to use.
According to the experts, IT departments everywhere may soon need to deploy protections akin to those now used by finance companies, which have struggled with fraud for centuries (see case study). In response, the once-staid authentication market is rapidly transforming. New startups, new form factors, and an influx of venture capital mean help is on the way. The challenge for enterprise IT is to make it all work before the fraudsters find their way in.
The problem with passwords
As Bob Blakley sees it, it’s not that passwords have outlived their usefulness. It’s just that they never really worked to begin with.
“The basic problem is that there’s a built-in trade-off between the human cognitive capability and password strength,” says Blakley, who is chief scientist for security and privacy at IBM. If standard strong-password protocol is to use values with eight or more characters and a mixture of alphabetic and numeric values, users either settle for passwords that aren’t secure or choose secure passwords they can’t remember.
Sam Tuohey, CTO of Stanford Federal Credit Union, reached the same conclusion in a more empirical fashion — an audit of password strength for 45,000 customers. Tuohey’s team threw standard cracking tools at the list of encrypted passwords and found that approximately 80 percent of the values could be cracked “in about a second,” Tuohey says.
For many years, simple passwords were a sufficient deterrent to relatively low levels of hacking and online crime. No longer. What was once acceptable laxness of user access is now an open invitation to sophisticated online criminals, who have quickly discovered how to make short work of passwords with phishing attacks, in combination with malicious code to harvest other sensitive data.
Changes in the threat environment are spurring rapid change in the authentication business, says Chris Young, senior vice president and general manager of consumer solutions at RSA Security.
“You’ve seen a movement from high-school kids who write viruses to organized criminal rings that are doing phishing and pharming and propagating Trojans that steal information purely for profit,” Young says.
Stanford Federal knows that only too well. The credit union is hardly a target like Bank of America or Wells Fargo, but phishers still found it late last year and used a sophisticated and targeted scam to try to compromise customer accounts, Tuohey says.
Taking advantage of the credit union’s connection to Stanford University, the phishers harvested thousands of publicly available stanford.edu addresses and sent phishing e-mails to them, spoofed to look as if they came from the credit union. Tuhoey only knows of four customers who responded to the e-mail messages and says he doesn’t believe that any accounts were actually compromised in the scam. But the incident was a wake-up call.
Factors against fraud
Strong authentication — using additional factors such as smart cards, one-time password generators, and USB tokens — has been the traditional weapon of choice for organizations worried about fraud, and it’s still a popular choice for many organizations. RSA claims to have 20,000 customers worldwide using its SecurID token. But strong authentication has always been pricey to deploy and maintain, and many users find them inconvenient, especially in the U.S.
That was the conclusion that Stanford Federal Credit Union reached, as well. “Sending out 45,000 tokens, then supporting them when people broke or lost them, would have been prohibitive,” Tuohey says.
Stanford Federal Credit Union does use smart cards for employees who travel and work from home, but traditional smart cards wouldn’t have been practical for customers who don’t own readers to insert them into, Tuohey says. The credit union’s solution was to turn to a friendlier form of two-factor authentication, including anti-fraud and Web site authentication technology from PassMark Security (now part of RSA).
PassMark’s technology employs user-selected watermarks to distinguish legitimate Web pages from phishing scams, plus back-end anti-fraud analytics that spot suspicious log-in attempts. RSA calls this multipronged effort “adaptive authentication,” but it is more commonly known as “risk-based authentication.”
“It’s the concept of different types of authentication based on context,” RSA’s Young says. “Who are you? Where are you in the session? What’s your typical account behavior?”
Among the factors that anti-fraud vendors consider are time of day, the IP address and kind of computer used, and geographic location. Although these measurements aren’t foolproof, they’re highly accurate in identifying most users, experts say.
“My wife always banks online at home,” says Nico Popp, vice president of authentication services at VeriSign, which bought fraud detection company Snapcentric in February. “She’s going to have a very stable cluster of behaviors: the same kind of browser, the same ISP, and she always banks on Saturday morning. It’s a very clear pattern.”
If Mrs. Popp tries to log in from Korea on a different machine with non-English-language settings, it should set off alarms, Popp says. On the other hand, Popp himself is more of a globetrotter; for his profile the geo-location information is less reliable. But he does always connect from the same laptop, so the device settings and session information is just as powerful, he says.
Proceed with confidence
The combination of fraud detection and risk-based authentication is powerful, Popp says, because it is invisible to users under normal circumstances but springs to life when the risk associated with user behavior increases, as in the case of money transfers or sudden account changes.
But risk-based authentication is no silver bullet for enterprises, notes Stu Vaeth, chief security officer of Diversinet, a supplier of token-based strong authentication solutions. Anti-fraud and risk-based authentication are great at weeding out phishing and man-in-the-middle attacks, he says, but they aren’t as secure as traditional two-factor authentication.
RSA’s Young concurs. “It’s kind of like saying that alarm systems will make door locks go away,” he says. “What this will do is allow millions of consumers or enterprise users who are not using credentials like SecurID to open up their protection options.”
That kind of thinking represents a major shift in the authentication market. Whereas at one time merely granting permission was seen as the essence of authentication, today’s solutions are moving instead toward an idea of “confidence,” IBM’s Blakley says.
“People think of authentication as something to do at the beginning of a session and never do again, but authentication is a confidence building thing — you have to have confidence in the identity of your transaction partner, and that confidence can erode over time,” Blakley says. For example, getting through the identity check at the front gate of NSA headquarters in Fort George Meade, Md., doesn’t necessarily give a visitor access to every room in the building, he adds.
At security vendor Cydelity, the idea is to monitor users’ behavior after they’re logged on and flag what’s risky, according to CEO Bob Ciccone. “Enterprises have typically deployed layered defenses, but there’s not a layer where they’re watching what users do once they’re in,” he says.
As do other companies in the anti-fraud space, Cydelity considers geo-location and atypical behavior, such as changing or disabling e-mail notification in conjunction with money transfer requests and attempts to access from suspicious locations.
Increasingly, customers are combining this kind of analytics-based risk detection with soft, two-factor alternatives to tokens and smart cards that are easier to deploy and support. For example, Diversinet’s soft tokens offer strong authentication akin to traditional tokens but can be delivered over a wireless network and stored on a PDA or mobile phone. Bharosa, meanwhile, offers a choice of form factors for its Authenticator soft two-factor authentication application, while on the back end its Tracker application monitors the origin of log-ins to avoid fraud. Metrics used include the computer or mobile device used to log in, geo-location, and behavioral profiles, says Bharosa CEO Jon Fisher.
A wide-open future
According to guidance from the Federal Financial Institutions Examination Council (FFIEC) in 2005, “Single-factor authentication, as the only control mechanism, is inadequate for high-risk transactions” such as money transfers.
In June, the White House’s Office of Management and Budget seconded that, directing federal agencies to comply with NIST security standards, including encryption of data on mobile devices and two-factor authentication for remote access to data.
These advisories have sent financial institutions and government agencies scrambling to shore up user authentication with additional factors. But that’s not necessarily a good thing for enterprises. With vendors focused on consumer fraud protection for the government and financial verticals, enterprise-targeted products are being put on hold.
“The market opportunity is such on FFIEC that right now we’re 99 percent on that,” VeriSign’s Popp says. “Between fraud, identity theft, and regulations, vendors are all-hands-on-deck.” But when the flood of FFIEC-compliance money dwindles, he says, companies will begin looking to tap the even larger enterprise authentication market.
Like Popp, IBM’s Blakley sees a role for risk-based analysis as part of the ordinary authentication process at organizations of all stripes. “Right now people mostly do risk analysis up front. It’s plausible that in the future you’re going to have more dynamic assessments of risk factors, so if a system becomes aware that something squirrelly is going on, you’re asked to pass an additional authentication test to increase confidence in the strength of the authentication,” he says.
Customers can already combine identity analytics with business rule checks to spot relationships within enterprise user populations. Adding more authentication data into that mix will lead to even more focused offerings, Blakley says.
But the future of strong authentication may lie outside the hands of any one vendor. The open source Initiative for Open AuTHentication (OATH) now boasts more than 66 members, including smart-card vendor Axalto, BMC, IBM, USB-token maker SanDisk, and VeriSign, among others. The idea is to create an ecosystem of authentication hardware and software that is based on open source components, encouraging creativity in a market that has long been dominated by a handful of large companies.
“One thing we’ve pushed with OATH is an open approach to fraud detection. Proprietary networks will never succeed, if each vendor says, ‘This is my fraud data, and I’m not going to share it.’ That just helps the bad guys,” VeriSign’s Popp says.