The news in government cybersecurity is not all bad.
Following a slip in compliance scores for IT security requirements in fiscal 2012, scores rebounded in FY 2013. And a new emphasis on continuous monitoring and authorization of IT systems – together with a program to provide the necessary tools for the job – could mean that things will get a little better when the results are in for the fiscal year just ended.
The overall state of government cybersecurity is judged by the Federal Information Security Management Act, and the scorecard is the Office of Management and Budget’s annual report to Congress on FISMA compliance. In the report for FY 2012, released in early 2013, overall FISMA compliance slipped from 75 percent in FY 2011 to 73 percent.
In the report for FY 2013 however, overall performance jumped to 81 percent, “with significant improvements in areas such as the adoption of automated configuration management, remote access authentication and email encryption.”
I am the first to admit that FISMA compliance – or compliance with any standards – does not equate to security. But the reports provide a useful baseline and indicate that agencies are paying attention to their security and the maturity of their programs.
Patrick Howard, former chief information security officer for the Nuclear Regulatory Commission and the Department of Housing and Urban Development (and now the program manager for continuous diagnostics and mitigation (CDM) at Kratos Defense), points out that the most recent results show that agencies still are struggling to develop long-term security plans, and he expects to see this again for FY 2014. “That’s nothing new,” he said. “We’ve been seeing that for years.”
But there are some reasons to believe – or at least hope – that there will be continued improvement. The latest report cited an improvement in meeting cross-agency performance goals, including trusted Internet connections, strong authentication and continuous monitoring. And there will be a stronger emphasis on continuous monitoring in the next evaluations.
In November 2013, OMB Memo M-14-03 set a timeline for agencies to move from static reauthorization of IT systems every three years, to continuous monitoring and ongoing reauthorization. Agencies were to have a strategy for information security continuous monitoring (ICSM) in place by Feb. 28, 2014, begin cooperation with the Homeland Security Department to implement the plans and begin procuring products and services through the DHS CDM program. Agencies will be evaluated on their compliance with these requirements in their 2014 FISMA reviews.
Challenges to fully implementing these ICSM goals remain, of course. DHS has not yet established a governmentwide ISCM dashboard, as called for in the memo. And the CDM program, which provides a source for procuring tools and services through a blanket purchase agreement at the General Services Administration, still is a work in progress.
Two of six task orders to be released under Phase 1 of CDM have been released for industry quotes, and the remaining four orders are expected to be released in fiscal 2015. Phase 2 of the CDM program still is being developed. Howard says there is a lack of awareness among many agencies about the continuous monitoring services available under CDM and that many agencies are waiting to see what happens with the second task order before implementing these services.
I am hopeful that the increased resources and attention on continuous monitoring – both in formal programs and in the security community in general – will help continue the upward trend in FISMA scores, however. Higher scores might not mean that agency IT systems are more secure, but they couldn’t hurt.
Posted by William Jackson on Oct 03, 2014 at 12:33 PM0 comments
We all know the gears of government grind slowly, but when it comes to the arcane world of government encryption standards, “slowly” can mean something else entirely. When government time meets technology time, sparks can fly.
Take SHA-1, for example. That 160-bit hash algorithm has been at the heart of vital web security protocols such as Secure Sockets Layer (SSL) and Transport Layer Security (TLS) since shortly after it was developed by the National Security Agency in the 1990s. It has also been a core member of the FIPS standards published by National Institute of Standards and Technology.
However, it’s been under fire for nearly a decade. In 2005, a professor in China demonstrated an attack that could be successfully launched against the SHA-1 hash function, a feat that led to a lot of soul searching within the encryption community. Less than a year later, NIST was urging agencies to begin moving away from SHA-1 and toward stronger algorithms.
At the beginning of 2011, NIST went even further and put what seemed the final kibosh on the beleaguered algorithm by stating definitely that “SHA-1 shall not be used for digital signature generation after December 31, 2013.”
But earlier this year, stories began to emerge pointing out that despite the NIST statement, many government entities were still generating new SSL certificates using SHA-1 in favor of stronger versions.
In a February survey, web services company Netcraft found that fully 98 percent of all the SSL certificates used on the web used SHA-1 signatures, and less than 2 percent used the 256-bit SHA-256. The company also pointed out that a huge number of those certificates as originally issued would still be valid beyond the start of 2017.
It’s not that the security provided by these certificates has so far proven to be porous, but a so-called collision attack could open up valid certificates to be substituted by ones constructed by attackers, allowing them to circumvent web browser verification checks.
It would be time-consuming and need a lot of computing power, but the increasingly market-driven nature of the threat industry is making that less of a barrier. Researchers have shown how the price of an SHA-1 attack will rapidly shrink over the next few years.
That’s all driving a sense of inevitability about the continuing use of SHA-1. Companies such as Microsoft and Google said some time ago they would start winding down the use of the algorithm in their products, and now the browser companies are getting on board.
The developers of Chrome, for example, recently said they will start sunsetting the use of SHA-1 beginning with a release due in November, and on Sept. 23 those in charge of Mozilla-based browsers such as Firefox said they also will be “proactively” phasing out their support of certificates that use SHA-1 signatures.
What’s a government agency to think of this? There have certainly been confusing signals along the way. In 2012, the year after it said it wanted agencies to move away from SHA-1, NIST announced the winner in a competition to create a secure hash algorithm that could eventually be the basis of a new federal SHA-3 standard.
But at the same time, NIST downplayed the need for a new standard in the shorter term, saying SHA-2 seemed to be working just fine (though NIST recent issued a request for comments on a new FIPS 202 that will validate the use of SHA-3). Meanwhile. the current version of NIST secure hash standards (FIPS PUB 180-4) still lists SHA-1 as valid for use in government applications. At the rate the private sector seems to be moving, however, it seems that will soon be impractical.
Posted by Brian Robinson on Sep 26, 2014 at 10:21 AM0 comments
It has been a brutal season for data breaches, from the wholesale theft of customer records numbering in the billions to the exposure of naughty celebrity pictures. More significant to agencies is the case that cost US Investigations Services (USIS) a contract to perform government background checks.
It was bad enough when USIS gained attention as the contractor that vetted NSA leaker Edward Snowden and Washington Navy Yard shooter Aaron Alexis. But in the wake of an IT breach that might have exposed the files of thousands of Homeland Security employees, the Office of Personnel Management in September said “enough,” and dropped the company.
The growing pressure by hackers against high value targets and the volumes of personal and other sensitive information being stolen highlights one of the basic questions of cybersecurity: How do you keep the bad guys out?
Identity management and access control are the front lines of security. The ability to accurately identify users and control what they do within your systems is what separates insiders from outsiders. It has been apparent for some time that the traditional tool for this task – the password – is inadequate for the job, and biometrics is emerging as an alternative.
Which is better? The answer is that neither is adequate for strong, practical security on its own. Each has strengths and weaknesses, and real security requires some combination of these or other technologies.
The password by itself actually is a pretty good tool. It is simple to use, easy to implement and can be reasonably strong. The problem is one of scale. For a user juggling passwords for multiple accounts and for administrators juggling many users, the system quickly becomes unwieldy, and strong security begins to break down. In addition, the steady growth in computing power erodes password security by making dictionary and brute force attacks more practical.
Biometrics – the use of physical traits such as fingerprints, irises, faces or voices to identify persons – is more complex, but is becoming more practical. It offers the promise of better security based on the premise that there is only one you.
Yet it has its drawbacks. All forms of biometrics operate on the “close enough” principle. Whereas a password must be exact to be accepted, matching a biometric trait requires a judgment about whether there is a proper match. This leaves room for mistakes, either as false positives or false negatives. The algorithms making the decision can be tuned depending on the level of security required. But higher security comes at a cost in the form of increased time or computing power to determine a match and by increasing the possibility that a legitimate biometric will be rejected. And although there is only one you, biometric systems can be susceptible to spoofing. A stolen digital template of a biometric trait could be inserted into the authentication process to authenticate the wrong user.
There are other ID management technologies, of course, such as digital certificates, a form of electronic ID vouched for by a trusted party. These can be powerful, but also challenging to manage on a large scale.
The bottom line is, no matter how much these technologies improve, no single tool is likely to be good enough for really practical strong authentication, and it is unlikely that a new and perfect technology will come along any time soon. None of these technologies is a complete failure, either. By combining strengths to offset weaknesses, these common tools can be integrated into multifactor authentication that provides security that is stronger than the sum of its parts.
Government already has a tool that can enable multi-factor authentication, the Personal Identity Verification Card and its military counterpart, the Common Access Card. Taking full advantage of these for access control could go a long way toward improving federal cybersecurity.
Posted by William Jackson on Sep 19, 2014 at 6:56 AM3 comments
There are many ways bad guys attack systems, disrupt infrastructures and steal data, but one of the most common uses an entry point that is vital to Internet communications and yet, it seems, carelessly disregarded: the humble, but crucial, SSL.
Secure Sockets Layer is the standard way of establishing an encrypted link between a server and a client, such as a web browser. All transactions that are needed for modern Internet-based communications and commerce – credit card numbers, personal identifiers such as Social Security numbers, website logins etc. – use SSL.
But despite the moniker, SSL is sometimes not that secure. One particular and apparently growing problem is with improper SSL validation. That was the focus of the GoTo bug discovered early this year (and since patched) in Apple’s iOS and Mac OS X. The vulnerability opened up users of those systems to so-called man-in-the-middle (MITM) attacks, in which those with a “trusted” certificate can insert themselves into a communication stream between systems and read its contents.
Similar concerns are being expressed about Android devices.
Given that there are now some 1.3 million apps in the Google Play store, a million or so of which are free, it would take a long, long time to test each of them to see if they are vulnerable to an MITM attack. Organizations, or individual users, can test a limited number of apps they use without much problem (such as with this method). Testing a wide range of apps to certify that they are OK for people to use is a different matter.
Fortunately, security organizations are starting to catch up with the need. In August, the CERT Coordination Center at Carnegie Mellon University, which works with many companies and government agencies, introduced CERT Tapioca (Transparent Proxy Capture Appliance), a virtual machine that automates MITM analysis.
According to CERT/CC researcher Will Dorman, Tapioca so far is only catching low-hanging fruit, but it at least doesn’t take up any of his time, and it’s already caught several hundred vulnerable applications in just a few weeks of use.
The issue should be getting even more press than it has, particularly in government circles, since there are expectations that Android devices could become more attractive in the public sector with the introduction and further development of Samsung’s Knox containerization technology. Apart from device-specific elements of Knox, which Samsung is keeping to itself, most of the technology could find itself incorporated in Google’s next generation operating system, Android L.
Samsung itself got some criticism late in 2013, when researchers from Ben Gurion University in Israel said they had found a vulnerability on a Galaxy S4 device that was using Knox, but Samsung later said that wasn’t a fault with Knox itself. The company also said Knox offers additional protection against MITM attacks through mobile device management and a feature that allows traffic only from designated and secure apps to be sent via VPN tunnel.
Altogether, it’s not been a good year for SSL. In April, a major vulnerability in OpenSSL, the so-called Heartbleed bug, was revealed, one that had been around for over a year before anyone noticed it. That was also fixed, but it’s still an ongoing concern. Researchers at IBM also recently reported that, though attacks using Heartbleed have quieted down, there might still be as many as 250,000 servers left unpatched.
The OpenSSL Project, pushed by the flak it got from the Heartbleed fiasco, for the first time recently published the policy of how it handles security issues. Internally, it says, it divides security issues into low, moderate and high severity and will notify the openssl-announce list and update the organization’s home page when any fixes are planned.
SSL users can also get help through a recently started SSL Blacklist, an online and downloadable resource of SSL certificates associated with malware or botnet activities.
None of the potential problems with SSL are all that new, but with attackers becoming ever more sophisticated in their methods, as is the malware they use to disrupt systems and extract sensitive data, at least those problems seem to be getting more attention, along with tools to address them.
Posted by Brian Robinson on Sep 12, 2014 at 12:12 PM0 comments