The government says it did not know about the Heartbleed vulnerability in OpenSSL before it was publicly disclosed. But White House Cybersecurity Coordinator Michael Daniel says that if it had known, it might not have told us.
“In the majority of cases, responsibly disclosing a newly discovered vulnerability is clearly in the national interest,” Daniel wrote in a recent White House blog post. But not always. “Disclosing a vulnerability can mean that we forego an opportunity to collect crucial intelligence that could thwart a terrorist attack, stop the theft of our nation’s intellectual property or even discover more dangerous vulnerabilities that are being used by hackers or other adversaries to exploit our networks.”
Daniel goes on to explain some of the criteria used in deciding when and when not to disclose a serious vulnerability.
Over the years, the security community has come to a consensus on how to handle disclosure of security vulnerabilities in software. The discoverer first informs the product’s vendor, giving the company time to develop a patch or workaround before reporting it publicly. This protocol is not mandatory, however. Researchers can use the threat of disclosure to pressure vendors to respond to vulnerabilities, and some companies offer a bounty for new vulnerabilities to encourage researchers to cooperate. But the value of a new vulnerability can be much greater than a bounty.
In the end, how a vulnerability is handled depends on the motives and morals of the discoverer. For criminals, a good zero-day vulnerability—one for which no fix yet exists—is money in the bank. For governments, it can be an espionage tool or a weapon. The Stuxnet worm, an offensive weapon widely believed to have been developed by the United States and Israel, exploited several zero-day vulnerabilities.
Daniel said there are practical limits on hoarding bugs. “Building up a huge stockpile of undisclosed vulnerabilities while leaving the Internet vulnerable and the American people unprotected would not be in our national security interest,” he wrote. “But that is not the same as arguing that we should completely forgo this tool as a way to conduct intelligence collection and better protect our country in the long-run.”
Daniel said there are no hard and fast rules for determining when to disclose, but the administration has a “disciplined, rigorous” process for deciding. The criteria include:
- How widely used and important is the vulnerable product?
- How serious is the vulnerability. Can it be patched, and how much harm could it do if it falls into the wrong hands?
- Would we know if someone else was using it?
- What is the value of the intelligence we could gather with it, and are there other ways to gather it?
- Is someone else likely to discover it?
Heartbleed potentially leaked sensitive information protected by OpenSSL, which is very widely used to protect online commerce and other transactions. The vulnerability was critical, and although a fixed version of the software was released, replacing it will take some time.
Would we know if someone was using it? Maybe. Gathering useful information requires a high number of connections to a vulnerable server, which could be detected in activity logs. Shortly after the disclosure, Canadian police arrested a young man for allegedly using Heartbleed to steal tax data.
As for the value of intelligence to be gained, who can say? Is someone else likely to discover it? Yes, given that it was in open source software available to anyone. And did someone else discover it? Yep. Researchers at Codenomicon and Google Security.
So, did the National Security Agency discover Heartbleed first, and if they did would they have told us? According to White House criteria, it would be a good candidate for disclosure. But we’ll probably never know.
Posted by William Jackson on May 16, 2014 at 9:03 AM1 comments
In the wake of major data breaches during the last few months, and with the ongoing scare over the OpenSSL Heartbleed bug, government security managers will likely find themselves dealing with major changes in some of the encryption tools they use to safeguard their agencies’ data.
One change is already in the works. The National Institute of Standards and Technology recently released an update to its Special Publication 800-52, which offers guidelines on how to implement transport layer security (TLS) protocols. TLS is the standard used to protect sensitive data — anything from credit card numbers to patient health information, email and social networking details — that have to move across open networks, such as the Internet.
The Internet Engineering Task Force (IETF), which oversees Internet standards, found vulnerabilities in the 1.0 version of TLS and spent several years upgrading it through versions 1.1 and 1.2. Meanwhile, NIST withdrew the initial 800-52 guidelines, which it released in 2005, because it didn’t include any of the updates included in TLS 1.1 and 1.2.
The latest revision of the guidelines rectifies that, and offers network administrators a range of recommendations on how to configure TLS options, according to NIST, including which algorithms to use and the length of cryptographic keys.
And admins will have to deal with even bigger changes down the road. The IETF is currently considering yet another update in the TLS spec, to 1.3, which will likely remove the RSA key transport cipher suites. The RSA code has been the decades-old basis of the “handshakes” that are used to control network sessions, and are considered to be too vulnerable to attack.
Improving the TLS handshake is one of the prime design targets for the IETF TLS 1.3 working group. The current IETF thinking is apparently to instead use systems such as the Diffie-Hellman Exchange or Elliptic Curve Diffie-Hellman Exchange. Both of those support perfect forward secrecy, which many organizations are pushing because they see it as offering much better protection for the keys used to encrypt data and communications. Perfect forward secrecy is set up when a compromise of one message cannot lead to the compromise of others.
However, perfect forward secrecy is not entirely perfect, as it turns out. All of the network sessions signed using a leaked or stolen key would still be open to compromise, for example. So while it would work for all future sessions it wouldn’t retroactively solve problems caused by the Heartbleed bug.
But the need for something better, and for the features TLS 1.3 will offer, is now much clearer. Heartbleed itself is still being examined, though early attempts to estimate the costs from fixing the many systems that could be affected suggest there will be major repercussions.
A more tangible indicator is what happened with the data breach at Target stores last year. Analysis of the numbers involved with that breach — which was not even the year’s biggest — are staggering: $200 million for banks to reissue cards to customers whose card numbers were stolen, and $100 million for Target to upgrade its payment terminals. The company also saw a 46 percent hit to its bottom line in the fourth quarter of 2013, as it struggled to regain its customers’ confidence.
There’s a takeaway from that analysis that NIST, the IETF or anybody else could have fixed, however. There were precisely zero people at Target with the title of either chief information security officer or chief security officer.
Posted by Brian Robinson on May 09, 2014 at 10:13 AM0 comments
The heartbleed vulnerability, which can leak sensitive data from supposedly secure Web connections, exposes the limits of using one-off credentials that must be authenticated separately for each transaction. Attack surfaces are greatly expanded when personally identifiable information (PII) is maintained by every agency and Web site offering online services.
“The idea that the user must have information everywhere is a bad idea,” said Andre Boysen, executive vice president for marketing at SecureKey. Having a single credential that can be authenticated by a trusted authority and accepted by multiple users can reduce the attack surface by maintaining PII at a single point. It also helps relieve the burden of managing credentials and identities.
This idea of federated identity is not new. Banks, merchants and credit card companies have been using a form of it for years. Merchants no longer have to issue and manage their own credit cards. A bank vets your identity and creditworthiness, a card company ensures the credit card is valid and has not been compromised, and online merchants do not have to worry about who you are as long as a credit card company vouches for the card.
It is not a risk-free system, but the risk is managed. Credit card numbers are sometimes exposed, but the exposure is considerably less than if every merchant had to maintain PII for every customer. When a breach occurs, users have to change one credit card, not one for every merchant visited.
Why can’t government online authentication be this simple? “It is heading that way,” Boysen said.
Canada implemented a Federated Identity Management program to leverage interoperable security credentials several years ago. In the United States, the Postal Service is preparing to roll out the Federal Cloud Credential Exchange (FCCX), a federated identity management hub that will let agencies accept online credentials issued by trusted third parties.
The system is part of the National Strategy for Trusted Identities in Cyberspace. Personal information and the identity of the original issuer of the credentials will be hidden from the FCCX hub, and log-in information will not be shared or compared between agencies. But the agency will know what it needs to know: You are who you say you are.
It should be noted that SecureKey is not exactly an impartial observer in this issue. The company has contracts with both Canada and USPS to provide a cloud-based platform for authenticating digital credentials. But this does not change the fact that a federated system offers a way to improve both security and privacy at a time when attacks on online activities are growing and digital credentials are like money in the bank for criminals.
Government stands to benefit greatly from federated identity schemes. Unlike banks, agencies tend to have relatively few transactions with each individual, which raises the overhead of authenticating each user that logs on. “Basically, every transaction is a re-enrollment,” Boysen said. This is frustrating to the user, expensive for the agency, and each agency also must manage and secure its own database of PII. Offloading authentication to a central hub eliminates the need to hold and protect that extra data.
No scheme will provide absolute security or completely transparent authentication. But federation and an interoperable system of trust can help. With such a system in place, agencies won’t have to worry if PII is leaking from their sites, and users would be able to whittle down the number of passwords and other credentials to be replaced when something does go wrong.
Posted by William Jackson on May 02, 2014 at 11:44 AM2 comments
Given the kinds of pressures it puts on organizations, taking care of security tends to be a short-term memory thing. It’s all about what happened yesterday and what’s likely to happen tomorrow, and even a year ago can seem like another reality.
Verizon, in its 2014 Data Breach Investigation Report (DBIR), shows that the short-term focus has its downsides. Sometimes, that longer-term view can produce real dividends that might otherwise be overlooked.
Foregoing the usual trends approach of these studies — what happened last year versus the previous one — Verizon instead took a deeper dive into the decade of data it has in its vaults.
The goal was to derive “actionable” information that would be of actual use. The report analyzed data from some 50 organizations around the world, taking a look at more than 63,000 security incidents that resulted in 1,367 confirmed data breaches in 2013. Previous DBIR’s had never had more than 1,000.
It’s over longer periods of time that the patterns show themselves, however. The attack methods that created the vast majority of those 2013 incidents can be reduced to just nine and, of those, three stand out as the main culprits. That also confirms the results Verizon found from analyzing previous years’ data.
Drill down even further and two classes of incidents — caused by Web app attacks and cyber espionage — are seen as the main breach threats over the past few years. Point-of-sale attacks, which had exploded through 2011, dropped off sharply in 2012 and only picked up again slightly in 2013. That’s due, Verizon analysts feel, to more small and medium sized businesses becoming aware of them.
“There’s been so many attacks of that kind over the last several years that it’s been like overfishing in the ocean,” said Marc Spitler, Verizon’s senior risk analyst. “So many of the SMBs have been caught, and they have now put meaningful defenses in place.”
The real revelation, however, comes with a comprehensive matrix of incident classifications and industries they’ve affected. That shows a huge difference in the kinds of hits that various sectors take.
It also produces surprises. While Web attacks and cyber espionage may be the strongest threats overall, as far as the public sector is concerned ‘insider misuse’ and ‘miscellaneous error’ accounted for nearly 60 percent of the entire number of incidents in 2013. The feared Web attacks and cyber espionage each accounted for less than 1 percent of public sector incidents.
Misuse and error can be anything from insiders deliberately stealing information to inadvertent mistakes, such as sending an email to the wrong person or attaching files that shouldn’t be attached, according to Spitler.
And that points to the kind of actionable information he hopes readers of the DBIR will glean from the report. Despite all of the fervor and money that’s being directed at government cyber security, a much more immediately effective remedy might be to really tighten up on information handling processes and procedures and general data hygiene?
The overall message of the Verizon report is probably summed up in a very simple graphic that compares, over 10 years, the time it has taken for an attacker to compromise an asset versus the time it takes for a defender to discover the breach. The gap is widening.
“We’re definitely not happy to see that trend,” Spitler said, “and it points to the fact that we need to see better detection methods and controls, because we need to pick these things up faster and sooner to prevent data loss.”
As a member of the self-confessed geek squad inside Verizon that spend a lot of their time going over incident and breach data, Spitler confesses he’s “extremely excited” by the more than 40 organizations who will likely add their data into the next DBIR, which can only improve the accuracy of the company’s big data analysis of these long-term security features.
“With better knowledge of these events we can do a better job of defending against the breaches they cause,” he said. “And it’s the little things we really need to understand.”
Posted by Brian Robinson on Apr 25, 2014 at 10:59 AM0 comments