WhatsApp, a popular cross platform messaging application, has been in the news recently, not only for the high-profile acquisition from Facebook, but also due to security concerns. A recently published article disclosed the WhatsApp database is exposed to third-party access and is not properly secured.

The Issues

The published analysis showed that WhatsApp can store both plain-text and encrypted databases on the SD card, where they could be accessed by malicious 3rd party applications. The only protection for the database is a simple AES cipher with an easily derived key, which leads to real security issues.

Because the database is encrypted as a whole, it’s very likely that at some point during application usage, the entire database itself must exist somewhere in a plain text form in order for the SQLite library to use it. This could pose a risk if the raw data became available to an attacker.

However, the most significant issue is that WhatsApp uses easily discoverable keys on the the encrypted database. Originally a static key was used to perform the encryption on all Android installations, though WhatsApp has now been updated to use a derived key based on the name of the email account on the device. In either case, this renders the encryption key available to applications running on the device, allowing exploits.

A Note on Key Management

Key management is a difficult issue, requiring careful consideration during application design. Storing the key directly within the application, or derived solely based on data on the device, is a poor practice. Decompiling a binary to access the key or determine the process used to create it is a rather straight forward process. Once the key has been identified, the data is at risk.

We strongly recommend that applications ensure that key data is provided by the user (e.g. a user supplied password). It’s fine to combine that with with device data to generate a password, but that is an additional precaution. It’s equally important to pass the key material through a key derivation function with a salt in to generate the actual key for encryption. This helps protect against dictionary and brute force attacks.

How SQLCipher Could Help

In the case of WhatsApp, SQLCipher can help address some of the factors above by providing a more secure environment for application data. Coupled with a user-supplied password or pin, SQLCipher’s built in security controls would automatically provide key derivation and encryption using a unique per-database salt. This would ensure that even if two databases are created with the same password they will not have the same encryption key. Furthermore, using SQLCipher, data is only decrypted as needed, decreasing the risk of plain-text data exposure. Finally, using SQLCipher would provide additional security features to ensure the integrity of the data, preventing tampering and closing off other potential attack vectors.

Many companies both small and large depend on SQLCipher to protect their data across a variety of platforms. Because WhatsApp is already using SQLite to store data, converting to use SQLCipher in its place would require minimal modifications, something WhatsApp might consider.

This Thursday night, February 27th at 9pm EST, Tempo and other web systems will be temporarily unavailable while we perform critical patch updates to ensure the stability of our services.

This maintenance outage will also affect the Tempo API, the the Connect website, and the site for Codebook.

Down time could last up to 1 hour, however we believe it will be completed much more quickly. If you need to get in touch with us for any reason, please don’t hesitate.

If you weren't already aware, Apple released iOS update 7.0.6 yesterday to patch a hole in their use of SSL/TLS that could allow an attacker to compromise encrypted communications. If you haven't updated your iOS devices you'll want to do so as soon as possible. A little bird tells me OS X updates are on the way next week, although I imagine that's no surprise and I've seen it speculated elsewhere.

There's a great explanation of the bug here by Adam Langley, but I'd like to point out an interesting comment by John Gruber on Daring Fireball where he links to Langley's post:

I don’t want to start a coding-style war, but I think this bug would not have happened if the code had been written using curly braces after the if statements.

I think there's no doubt about that. A little while ago I gave up my stubborn use of BSD-style in my C and Objective-C code and now wherever I see it I cringe a little and then go about fixing it up to be more "proper" in my mind. It's a nuisance, but worth it. For some time now, I think perhaps at the recommendation of Zed Shaw in his online book Learn C the Hard Way, I've been making sure to always use braces to encapsulate anything depending on an if statement or other conditional to avoid scenarios like the one that caused Apple's exploit. Other rules I now keep are no more braces on the next line, and avoiding blank lines. This gist should give a good example of what I mean, even though it's a bit silly:

I even use braces to encapsulate the branches of switch statements these days! I like to zap blank lines because I think it forces me to review code more closely; it's harder to skim and encourages me to keep things nice and DRY.

A minor aside: Learn C the Hard Way is really great even if you know C already. Worth a couple of hours, it's a fun read and good for tightening up your skills.

Update: I also avoid assignment operators in if statements, and I tend to keep conditional expressions very explicit. This way I know what in my code above, self.fugeez is a BOOL, or returns that type. If I want to check that a pointer is nil, then if (self.fugeez == nil), not if (self.fugeez). If I want to test that my BOOL fugeez is false, then if (self.fugeez == NO), an explicit comparison. Just makes it easier to see what's going on later. I'm not quite as religious about this, but I'm trying to stick to it and adjust where I see my old ways.

As a follow-up to the original Zetetic.Security library that adds PBKDF2 and BCrypt-based HashAlgorithms to .NET, we're pleased to announce the release of Zetetic.Security 1.1.0.0, which provides options for greater work factors when generating password hashes.

The original hash implementations use 5,000 rounds of PBKDF2, and a BCrypt work factor of 10^10.  The 1.1 release adds support for 64,000, 128,000, and 256,000 rounds of PBKDF2, and BCrypt work factors from 10^11 to 10^16.  As before, the best way to use these hash algorithms is by installing Zetetic.Security to the Global Assembly Cache, and updating machine.config to register the new classes.  (Hint: you can find the path to the effective machine.config file in System.Runtime.InteropServices.RuntimeEnvironment.SystemConfigurationFile).

Once installed and registered, the new HashAlgorithms can be used in code as per the following sample (I wrote this one in LinqPad for simplicity's sake):

Of course, bear in mind two important caveats before updating your applications to use any new password hashing strategies:

1. .NET MembershipProviders don't have any way to know what HashAlgorithm generated any particular user's hashed password in storage.  Therefore, changing the hash algorithm will cause your users' logins to fail until they reset their passwords.

2. A stronger hashing algorithm is likely to eat up significantly more CPU cycles.  Plan accordingly so that applying 10^16 rounds of BCrypt doesn't open the door to a denial-of-service attack (or blow your AWS or Azure budget!).

Download from Github

View on NuGet

At our STRIP meeting today we discussed providing more regular, up-to-date information about what versions of STRIP are currently available and what's on the way, a sort of status page. We haven't decided exactly how we'll implement this but for now we'd like to provide an update here on the blog.

STRIP for iOS 2.2.0 - Released - Jan 22, 2014

Minimum iOS version: 6.0

• Features a new interface customized for iOS 7
• Various bug fixes, minor feature improvements
• Sync-compatible with 2.1.0
• Crashes on sync with Google Drive on 64-bit devices

STRIP for Android 2.2.0 - Released - Jan 22, 2014

Minimum Android version: 2.1

• Various bug fixes, minor feature improvements
• Sync-compatible with 2.1.0

STRIP for Windows 2.1.0 - Released - Nov 12, 2013

Minimum Windows version: Windows XP

• Updates encryption engine to SQLCipher 3
• Increases PBKDF2 key derivation from 4,000 to 64,000
• Requires min version 2.1.0 for sync

STRIP for OS X 2.1.0 - Released - Nov 12, 2013

Minimum OS X version: 10.6 Snow Leopard

• Updates encryption engine to SQLCipher 3
• Increases PBKDF2 key derivation from 4,000 to 64,000
• Requires min version 2.1.0 for sync

STRIP for iOS 2.2.1 - Submitted

Minimum iOS version: 6.0

• Awaiting review from Apple
• Fixes crash on 64-bit devices during Google Drive sync