Two years ago, ElcomSoft analyzed some 17 password management applications for mobile platforms only to discover that no single app was able to deliver the claimed level of protection. The majority of the apps relied upon proprietary encryption models rather than utilizing iOS exemplary security model. As a result, most applications were either plain insecure or provided insufficient security levels, allowing a competent intruder to break into the encrypted data in a matter of hours, if not minutes. Full report (PDF) is available here.
Today, we need stronger security more than ever. Was the urge for stronger security recognized by software makers, or are they still using the same inefficient techniques? In order to find out, we decided to re-test some of the previously analyzed products. Keeper® Password Manager & Digital Vault will the first subject for dissection.
Back in 2012, we weren’t much impressed by security in any of the apps we analyzed. Two years later, Keeper developers claimed they’ve successfully implemented the suggestions we made during the last analysis. The developers claim to have used 256-bit AES encryption, PBKDF2 key generation, BCrypt, and SHA-1 among other things. Let’s see if these improvements lead to stronger security.
Three and a half years ago (in April 2009) our company took part in InfoSecurity Europe in London. I should confess that London is one of my favourite cities; besides, I love events on security — so that I was really enjoying that trip (with my colleagues). But something happened.
BitLocker, PGP and TrueCrypt set industry standard in the area of whole-disk and partition encryption. All three tools provide strong, reliable protection, and offer a perfect implementation of strong crypto.
Normally, information stored in any of these containers is impossible to retrieve without knowing the original plain-text password protecting the encrypted volume. The very nature of these crypto containers suggests that their target audience is likely to select long, complex passwords that won’t be easy to guess or brute-force. And this is exactly the weakness we’ve targeted in our new product: Elcomsoft Forensic Disk Decryptor.
The Weakness of Crypto Containers
The main and only weakness of crypto containers is human factor. Weak passwords aside, encrypted volumes must be mounted for the user to have on-the-fly access to encrypted data. No one likes typing their long, complex passwords every time they need to read or write a file. As a result, keys used to encrypt and decrypt data that’s being written or read from protected volumes are kept readily accessible in the computer’s operating memory. Obviously, what’s kept readily accessible can be retrieved near instantly by a third-party tool. Such as Elcomsoft Forensic Disk Decryptor.
Retrieving Decryption Keys
In order to access the content of encrypted containers, we must retrieve the appropriate decryption keys. Elcomsoft Forensic Disk Decryptor can obtain these keys from memory dumps captured with one of the many forensic tools or acquired during a FireWire attack. If the computer is off, Elcomsoft Forensic Disk Decryptor can retrieve decryption keys from a hibernation file. It’s important that encrypted volumes are mounted at the time a memory dump is obtained or the PC goes to sleep; otherwise, the decryption keys are destroyed and the content of encrypted volumes cannot be decrypted without knowing the original plain-text password.
“The new product includes algorithms allowing us to analyze dumps of computers’ volatile memory, locating areas that contain the decryption keys. Sometimes the keys are discovered by analyzing byte sequences, and sometimes by examining crypto containers’ internal structures. When searching for PGP keys, the user can significantly speed up the process if the exact encryption algorithm is known.”
It is essential to note that Elcomsoft Forensic Disk Decryptor extracts all the keys from a memory dump at once, so if there is more than one crypto container in the system, there is no need to re-process the memory dump.
The FireWire attack method is based on a known security issue that impacts FireWire / i.LINK / IEEE 1394 links. One can take direct control of a PC or laptop operating memory (RAM) by connecting through a FireWire. After that, grabbing a full memory dump takes only a few minutes. What made it possible is a feature of the original FireWide/IEEE 1394 specification allowing unrestricted access to PC’s physical memory for external FireWire devices. Direct Memory Access (DMA) is used to provide that access. As this is DMA, the exploit is going to work regardless of whether the target PC is locked or even logged on. There’s no way to protect a PC against this threat except explicitly disabling FireWire drivers. The vulnerability exists for as long as the system is running. There are many free tools available to carry on this attack, so Elcomsoft Forensic Disk Decryptor does not include a module to perform one.
If the computer is turned off, there are still chances that the decryption keys can be retrieved from the computer’s hibernation file. Elcomsoft Forensic Disk Decryptor comes with a module analyzing hibernation files and retrieving decryption keys to protected volumes.
Complete Decryption and On-the-Fly Access
With decryption keys handy, Elcomsoft Forensic Disk Decryptor can go ahead and unlock the protected disks. There are two different modes available. In complete decryption mode, the product will decrypt everything stored in the container, including any hidden volumes. This mode is useful for collecting the most evidence, time permitting.
In real-time access mode, Elcomsoft Forensic Disk Decryptor mounts encrypted containers as drive letters, enabling quick random access to encrypted data. In this mode files are decrypted on-the-fly at the time they are read from the disk. Real-time access comes handy when investigators are short on time (which is almost always the case).
We are also adding True Crypt and Bitlocker To Go plugins to Elcomsoft Distributed Password Recovery, enabling the product to attack plain-text passwords protecting the encrypted containers with a range of advanced attacks including dictionary, mask and permutation attacks in addition to brute-force.
The unique feature of Elcomsoft Forensic Disk Decryptor is the ability to mount encrypted disks as a drive letter, using any and all forensic tools to quickly access the data. This may not seem secure, and may not be allowed by some policies, but sometimes the speed and convenience is everything. When you don’t have the time to spend hours decrypting the entire crypto container, simply mount the disk and run your analysis tools for quick results!
Elcomsoft has announced that certain versions of fingerprint software named Protector Suite made by UPEK (now part of Authentec) stores your Windows password in a ‘scrambled’ format in registry. This allows an attacker through different entry points to get easy access to a users Windows password. I have no reason not to believe Elcomsoft in their claims, but UPEK/Autentec seriously disagrees. In the middle of this I happen to have some questions, and an opinion regarding biometric software today.
I have lost count of all the times colleagues have approached me with a big smile, challenging me to break into their work laptops now that they have enabled fingerprint authentication. Pressing Esc to get the normal logon prompt and then entering my AD username & password logged me in. Having local admin rights made things even easier to conduct pass-the-hash of their locally cached credentials, and smile turned to sadness. Hey, I have even been accused of cheating when I did that.
I purchased my first fingerprint reader back somewhere in 1999. It was complete crap. Many years later I purchased a Microsoft keyboard with integrated fingerprint reader:
I still remember a very clear warning in their documentation: the fingerprint reader should not be trusted for security. It should be considered as a toy. Oh well.
Today the integrated fingerprint readers in many laptops is the most common place we interact with biometric solutions. IF we choose to use it of course – there is no requirement to do so from the vendor. Enter Elcomsoft.
Security vs Convenience
Lots of people – including infosec professionals, doesn’t see the difference between using biometric authentication as a security feature, and as a convenience feature. Simply explained for the home user:
If you use biometric authentication to logon to your laptop, but can bypass it by pressing Esc and enter your username & password, you are using biometrics as a convenience feature.
If you have removed any and all possibilities to logon except by using/including biometrics, you are using biometrics as a security feature.
The differences here are … well… BIG, at least in theory. But wait; that was for the home user. I don’t care much about your private pictures, christmas wish list and facebook account anyway, so lets look at it from a corporate perspective:
There is no integrated support for replacing passwords with biometric authentication within Microsoft Windows.
This means that any kind of authentication addition or replacement you set up on laptops, tablets or desktop computers in a corporate enviroment with Active Directory, a password still has to be configured for a user in a domain, and that password is what authenticates the user throughout the domain. Using highly advanced visualization tools, hours and hours of hard work and a colorful palette, I made this infographic to explain what happens:
Using biometric logon, we add another step in the authentication process in a corporate environment. Please note; we added one more step, we didn’t necessarily add one more layer of security.
I blogged about upcoming password security features in Windows 8 Password Security. Please observe that using picture password and/or a PIN is an addition to having a password. They are quite simply convenience features. Having said that, I would like to give kudos to Microsoft for doing quite a bit of research into picture passwords and presenting it in such a detailed form that we can make up an opinion about the security it provides.
What did Elcomsoft discover?
Well, they claim that certain versions of the software in question stores your Windows password using weak protection locally (see step 2 in the biometric chain above). Using a simple PoC, they have successfully extracted the stored Windows password from registry by the biometric software and “decrypted” it.
Since the biometric software is local only, it needs to know your Windows password to properly give you both local and domain access. To repeat; your username and password gives you access, not your fingerprint or any other biometric ID. If your password is changed, either locally or in the domain, you will have to provide your new password to the biometric software.
Is this such a big deal? Yes.
Good practice is to store passwords using hash irreversible algorithms, preferably strong types such as PBKDF2, Bcrypt or Scrypt. The draft cheat sheet from OWASP on password storage gives more information about such algorithms, and more. Even though Microsoft doesn’t use salting or key stretching in their LM/NTLM algorithms, they are still hash algorithms. You cannot “reverse” the process to get the plaintext password, you have to
My Authentec (Thinkpad) fingerprint software, which is NOT affected by Elcomsofts findings, knows my password (or passphrase in my case), and there is an option in the software to display it on screen, as the video on top shows you.
But I can do pass-the-hash/ticket and more, why is this a big deal?
Sure you can. But you cannot do those attacks against a Outlook Web Access configuration from the Internet using SSL. You don’t know the users actual password when you do pass-the-hash attacks, so you cannot check if the user uses the same password on other services, at work or on a personal basis.
If my fingerprint – my biometric template – was the secret key to unlock the password using reversible encryption like AES, things could perhaps be considered a bit better, but it would still not be good practice to store any users password using reversible encryption. Which is exactly what is evidenced by my video above.
Now if claims by Elcomsoft are true, malware could easily exploit the weakness found to extract users Windows plaintext passwords in yet another way, adding to the already existing ways of doing so.
I haven’t twisted my mind long enough on this to figure out ways of improving this, but I am open for suggestions.
ElcomSoft has recently updated two products recovering Microsoft Office passwords with Office 2013 support. Elcomsoft Advanced Office Password Recovery and Elcomsoft Distributed Password Recovery received the ability to recover plain-text passwords used to encrypt documents in Microsoft Office 2013 format. Initially, we are releasing a CPU-only implementation, with support for additional hardware accelerators such as ATI and NVIDIA video cards scheduled for a later date.
In version 2013, Microsoft used an even tighter encryption compared to the already strong Office 2010. To further strengthen the protection, Microsoft replaced SHA1 algorithm used for calculating hash values with a stronger and slower SHA512. In addition, the encryption key is now 256 bits long, while the previous versions of Microsoft Office were using ‘only’ 128 bits. While the length of the encryption key has no direct effect on the speed of password recovery, the slower and stronger hash calculation algorithm does. It’s obvious that Microsoft is dedicated to making subsequent Office releases more and more secure.
No Brute Force
While we continue supporting brute force attacks, brute force becomes less and less efficient with every new release of Microsoft Office even with full-blown hardware acceleration in place. Office 2013 sets a new standard in document encryption, pretty much taking brute force out of the question. This is why we continue relying on a variety of smart attacks that include a combination of dictionary attacks, masks and advanced permutations. Brute-forcing SHA512 hashes with 256-bit encryption key is a dead end. Smart password attacks are pretty much the only way to go with Office 2013.
Most laptops today ship with a fingerprint reader. Most likely, you have a laptop with one. Until very recently, most major manufacturers such as Acer, ASUS, Dell, Gateway, Lenovo, MSI, NEC, Samsung, SONY, Toshiba, and many others were using fingerprint readers manufactured by a single company: UPEK.
ElcomSoft discovered a major flaw with UPEK Protector Suite, which was the software shipped with the majority of laptops equipped with UPEK fingerprint readers until the company was acquired by Authentec and switched to different software. Even today, when UPEK is acquired by Authentec which now uses TrueSuite® software, many (or most) existing laptop users will simply stay with the old flawed software, not feeling the need to upgrade.
Does Fingerprinting the User Lead to Tighter Security?
Laptops normally come loaded with pre-installed software. Among other things manufacturers install on your brand-new laptop is software communicating with UPEK readers: UPEK Protector Suite. The suite manages fingerprint reading hardware, offering users the convenience of substituting the typing of passwords with a single swipe of a finger. Ultimately, UPEK Protector Suite caches your passwords, offering near-instant login to Web sites and Windows itself.
Logging into Windows by swiping a finger instead of clicking and typing a (probably long and complex) password sounds tempting. And, it works. A simple swipe of your finger, and you’re in. Wonderful; but what about security?
Here’s what UPEK says on its Web site about the Windows login: “Protector Suite QL allows for secure access to Windows by swiping your finger instead of typing a password.” Notice the “secure” part? Well, we found out UPEK makes Windows login anything but secure. In fact, the UPEK’s implementation is nothing but a big, glowing security hole compromising (and effectively destroying) the entire security model of Windows accounts.
The Issue with UPEK Protector Suite
After analyzing a number of laptops equipped with UPEK fingerprint readers and running UPEK Protector Suite, we found that your Windows account passwords are stored in Windows registry almost in plain text, barely scrambled but not encrypted. Having physical access to a laptop running UPEK Protector Suite, we could extract passwords to all user accounts with fingerprint-enabled logon. Putting things into perspective: Windows itself never stores account passwords unless you enable “automatic login”, which is discouraged by Microsoft. If you use the Windows auto-logon feature, you’ll see a message saying “Using automatic logon can pose a security risk because anyone that has access to your computer will have access to your programs and personal files.” Simply said, no corporate user will ever use this “automatic logon” feature, which is often banned by corporate security policies.
However, fingerprint logon is rarely, if ever, barred. The common perception is that biometric logon is just as, or maybe more secure than password-based one. While biometric logon could be implemented that way, UPEK apparently failed. Instead of using a proper technique, they preferred the easy route: UPEK Protector Suite simply stores the original password to Windows account, making it possible for an intruder to obtain one.
Storing Windows account passwords in plain text is bad practice. It defeats the entire purpose of enhanced security. In fact, with current implementation, we cannot speak of any security as the entire PC becomes extremely easy to exploit to anyone aware of this vulnerability. This time around, UPEK made it completely wrong, introducing a paper link to a stainless steel chain.
If Your Windows Logon Password Is Compromised
What happens if someone gets to know your Windows account password? First, they obviously gain access to all your files and documents. Of course, if they had your laptop and its hard drive at their disposal, they could to that anyway – with one exception: they would not be able to read EFS-encrypted files (those that have the “Encrypt contents to secure data” checkbox ticked in the file properties – Attributes – Advanced). EFS encryption is extremely strong and impossible to break without knowing the original Windows account password.
And here comes UPEK Protector Suite. Conveniently storing your plain-text account password, the suite gives the intruder the ability to access your used-to-be-protected EFS encrypted files. Bummer.
The Scope of the Issue
The scope of this issue is extremely broad. It is not limited to a certain laptop model or manufacturer. All laptops equipped with UPEK fingerprint readers and running UPEK Protector Suite are susceptible. If you ever registered your fingerprints with UPEK Protector Suite for accelerated Windows logon and typed your account password there, you are at risk.
Course of Action
If you care about security of your Windows account, launch UPEK Protector Suite and disable the Windows logon feature. That should clear the stored password for your account. Note that you should clear all stored account passwords to protect all user accounts.
What We Did
ElcomSoft will not disclose full detail in the interests of public responsibility. We notified former UPEK about the issue (but sure enough they know about it). We also prepared a demo application, which displays partial login credentials of users who enabled fingerprint login. We won’t give it away to general public; only a limited number of hi-tech journalists will receive this software.
Few days ago we have updated our iOS Forensic Toolkit to version 1.15 which includes some bugfixes and improvements and, most notably, supports passcode recovery on the new iPad (also known as iPad 3). There are no significant changes from the practical point of view (i.e. the process of passcode recovery is still exactly the same), but there is something new under the hood. So if you’re interested in iOS security and how stuff works, please read on.
It’s been almost two weeks since we have released updated version of Elcomsoft Phone Password Breaker that is capable of downloading backups from the iCloud and we have seen very diverse feedback ever since. Reading through some articles or forum threads it became quite evident that many just do not understand what we have actually done and what are the implications. So I am taking another try to clarify things.
We’ve analyzed 17 popular password management apps available for Apple iOS and BlackBerry platforms, including free and commercially available tools, and discovered that no single password keeper app provides a claimed level of protection. None of the password keepers except one are utilizing iOS or BlackBerry existing security model, relying on their own implementation of data encryption. ElcomSoft research shows that those implementations fail to provide an adequate level of protection, allowing an attacker to recover encrypted information in less than a day if user-selectable Master Password is 10 to 14 digits long.
Both platforms being analyzed, BlackBerry and Apple iOS, feature comprehensive data security mechanisms built-in. Exact level of security varies depending on which version of Apple iOS is used or how BlackBerry users treat memory card encryption. However, in general, the level of protection provided by each respective platform is adequate if users follow general precautions.
The same cannot be said about most password management apps ElcomSoft analyzed. Only one password management app for the iOS platform, DataVault Password Manager, stores passwords in secure iOS-encrypted keychain. This level of protection is good enough by itself; however, that app provides little extra protection above iOS default levels. Skipping the complex math (which is available in the original whitepaper), information stored in 10 out of 17 password keepers can be recovered in a day – guaranteed if user-selectable master password is 10 to 14 digits long, depending on application. What about the other seven keepers? Passwords stored in them can be recovered instantly because passwords are either stored unencrypted, are encrypted with a fixed password, or are simply misusing cryptography.
Interestingly, BlackBerry Password Keeper and Wallet 1.0 and 1.2 offer very little protection on top of BlackBerry device password. Once the device password is known, master password(s) for Wallet and/or Password Keeper can be recovered with relative ease.
Many password management apps offered on the market do not provide adequate level of security. ElcomSoft strongly encourages users not to rely on their advertised security, but rather use iOS or BlackBerry built-in security features.
In order to keep their data safe, Apple users should set up a passcode and a really complex backup password. The unlocked device should not be plugged to non-trusted computers to prevent creation of pairing. Unencrypted backups should not be created.
BlackBerry users should set up a device password and make sure media card encryption is off or set to “Encrypt using Device Key” or “Encrypt using Device Key and Device Password” in order to prevent attackers from recovering device password based on what’s stored on the media card. Unencrypted device backups should not be created.
Attacking Wi-Fi passwords is near hopeless if a wireless hotspot is properly secured. Today’s wireless security algorithms such as WPA are using cryptographically sound encryption with long passwords. The standard enforces the use of passwords that are at least 8 characters long. Encryption used to protect wireless communications is tough and very slow to break. Brute-forcing WPA/WPA2 PSK passwords remains a hopeless enterprise even if a horde of GPU’s is employed. Which is, in general, good for security – but may as well inspire a false sense of security if a weak, easy to guess password is selected.
Elcomsoft Wireless Security Auditor is one tool to test how strong the company’s Wi-Fi passwords are. After checking the obvious vulnerabilities such as open wireless access points and the use of obsolete WEP encryption, system administrators will use Wireless Security Auditor that tries to ‘guess’ passwords protecting the company’s wireless traffic. In previous versions, the guessing was limited to certain dictionary attacks with permutations. The new version gets smarter, employing most of the same guessing techniques that are likely to be used by an intruder.
Humans are the weakest link in wireless security. Selecting a weak, easy to guess password easily overcomes all the benefits provided by extensive security measures implemented in WPA/WPA2 protection. In many companies, employees are likely to choose simple, easy to remember passwords, thus compromising their entire corporate network.
The New Attacks
The new attacks help Elcomsoft Wireless Security Auditor recover weak passwords, revealing existing weaknesses and vulnerabilities in companies’ wireless network infrastructure.
If it’s known that a password consists of a certain word, the Word attack will attempt to recover that password by trying heavily modified versions of that word. This attack only has two options: you can set the source word and you can disable all permutations except changing the letter case. In addition, we can apply permutations to the source word first, forming a small dictionary; then perform a full dictionary attack, applying various permutations to all words from the newly formed list.
Certain passwords or password ranges may be known. The mask attack allows creating a flexible mask, brute-forcing the resulting limited combination of passwords very quickly. The masks can be very flexible. One can specify placeholders for static characters, letter case, as well as full or limited range of special characters, digits or letters. Think of the Mask attack as an easy (and very flexible) way to check all obvious passwords from Password000 to Password999.
You have two dictionaries. We combine each word from one dictionary with every word from another. By default, the words are combined as is, but you can increase the number of possible combinations by allowing delimiters (such as space, underscore and other signs), checking upper/lower case combinations or using extra mutations.
This is one of the more interesting attacks out there. In a sense, Hybrid attacks come very close to how real human intruders think. The Hybrid attacks integrates ElcomSoft’s experience in dealing with password recovery. We’ve seen many (think thousands) weak passwords, and were able to generalize ways people are making them. Dates, names, dictionary words, phrases and simple character substitutions are the most common things folks do to make their passwords ‘hard to guess’. The new Hybrid attack will handle the ‘hard’ part.
Technically, the Hybrid attack uses one or more dictionaries with common words, and one or more .rul files specifying mutation rules. We’re supplying a few files with the most commonly used mutation rules:
Common.rul – integrates the most commonly used mutations. In a word, we’ve seen those types of passwords a lot, so we were able to generalize and derive these rules. Dates.rul – pretty much what it says. Combines dictionary words with dates in various formats. This is a pretty common way to construct weak passwords. L33t.rul – the “leet” lingo. Uses various combinations of ASCII characters to replace Latin letters. C001 hackers make super-strong passwords with these… It takes minutes to try them all. Numbers.rul – mixes dictionary words with various number combinations.