The CrySL Language
Thanks to Theofilos Petsios from Amazon Web Services for providing a definition file for syntax highlighting for CrySL in VIM. You can download the definitions here.
Both CogniCryptGEN and CogniCryptSAST are based on CrySL rules that specify the correct use of an application programming interface (API). CrySL is a domain-specific language that allows to specify usage patterns of APIs. CogniCryptGEN generates code using the rules, CogniCryptSAST in turn reports any deviations from the usage pattern defined within the rules.
Syntax of the Domain-Specific Language CrySL
Rules in CogniCrypt are written in CrySL. CrySL is a domain-specific language for the specification of correct cryptography API uses in Java. The Eclipse plugin CogniCrypt ships with an XText editor that supports the CrySL syntax. CrySL generally encodes a white-list approach and specifies how to correctly use crypto APIs. We discuss some of the most important concepts of the rule language here, the research paper provides more detailed insights on the language. CogniCrypt ships with a default rule set for the Java Cryptographic Architecture (JCA). At the bottom of this page, you find a description of this rule set. On top of this rule set, rule sets for BouncyCastle, both for its lightweight API as well as JCA provider, and Google Tink are available for download from within the CogniCrypt preferences. Custom rules may also be added.
Each CrySL rule is a specification of a single Java class. A short example of a CrySL rule for javax.crypto.Cipher
is shown below.
SPEC javax.crypto.Cipher
OBJECTS
java.lang.String trans;
byte[] plainText;
java.security.Key key;
byte[] cipherText;
EVENTS
Get: getInstance(trans);
Init: init(encmode, key);
doFinal: cipherText = doFinal(plainText);
ORDER
Get, Init, (doFinal)+
CONSTRAINTS
encmode in {1,2,3,4};
alg(trans) in {"AES", ..., "RSA"};
alg(trans) in {"AES"} => mode(trans) in {"CBC"};
REQUIRES
generatedKey[key, part(0, "/", trans)];
ENSURES
encrypted[cipherText, plainText];
Each rule has a SPEC
clause that lists the fully qualified class name the following specification holds for (in this case javax.crypto.Cipher
)
The SPEC
clause is followed by the blocks OBJECTS
, EVENTS
, ORDER
, CONSTRAINTS
, REQUIRES
and ENSURES
.
Within the CONSTRAINTS
block each rule lists Integer
and String
constraints. The OBJECTS
clause lists all variable names that can be used within all blocks of the rule. The EVENTS
block lists API method calls that can be made on each Cipher
object. When an event is encountered, the actual values of the events parameters are assigned to respective variable name listed in the rule. These parameter values can then be constrained by CONSTRAINTS
.
The CONSTRAINTS section
The Cipher
rule lists encmode in {1,2,3,4};
within its CONSTRAINTS
block. The value encmode
that is passed to method init(encmode, cert)
is restricted to be one of the four integers. In other terms, whenever the function init
is called, the value passed in as first parameter must be in the respective set. The constraint alg(trans) in {"AES", ..., "RSA"}
refers to the fact that at the call to Cipher.getInstance(trans)
the String trans
must be correctly formed. Hence the constraint restricts the algorithm to be either "AES"
or "RSA"
through the alg
function. The third constraint (alg(trans) in {"AES"} => mode(trans) in {"CBC"};
) is a conditional constraint: If the algorithm of trans
is "AES"
, then the mode of trans
must be "CBC"
. For example, this conditional rule warns a developer writing Cipher.getInstance("AES/ECB/PKCS5Padding")
instead of Cipher.getInstance("AES/CBC/PKCS5Padding")
.
The ORDER section
The ORDER
section of a rule specifies a regular-expression like description of the expected events to occur on each individual object. For the Cipher
rule the order is Get, Init, (doFinal)+
. The terms Get
, Init
and doFinal
are labels and group a set of API methods that are defined within the EVENTS
block. The regular expression stated in the ORDER
section enforces the following order on a Cipher
object: The object must be create by a Get
, i.e., Cipher.getInstance
, call, then Init
must be called before, eventually, the method doFinal
is called. A programmer who writes the program below contradicts the ORDER
section of the CrySL
rule: A call to init
on the cipher
object is missing between the getInstance
and doFinal
call (the missing call is commented out).
Cipher cipher = Cipher.getInstance("AES/ECB/PKCS5Padding");
//cipher.init(Cipher.ENCRYPT_MODE, secretKeySpec);
cipher.doFinal(plainText);
The ENSURES and the REQUIRES section
Cryptographic tasks are more complex and involve interaction of multiple object instances at runtime. For example for an encryption task with a Cipher
instance, the Cipher
object must be initialized with a securely generated key. The API of the Cipher
object has a method init(encmode,key)
where the second parameter is the respective key and is of type SecretKeySpec
. For a correct use of the Cipher
object, the key must be used correctly as well.
To cope with these object interactions, CrySL allows the specification of what we call predicates that establish a rely-guarantee mechanism. Predicates are listed in the blocks REQUIRES
and ENSURES
. An object that is used in coherence with the rule receives the predicate listed in the ENSURES
block. In turn, the block REQUIRES
allows rules to force other objects to hold certain predicates.
The specification of the Cipher
rule lists a predicate generatedKey[key,...]
within its REQUIRES
block. The variable name key
refers to the same object that is used within the event Init: init(encmode, key);
of the EVENTS
block. Hence, the key object must receive this predicate which is listed in the rule for javax.crypto.SecretKeySpec
.
SPEC javax.crypto.spec.SecretKeySpec
OBJECTS
java.lang.String alg;
byte[] keyMaterial;
EVENTS
c1: SecretKeySpec(keyMaterial, alg);
...
REQUIRES
randomized[keyMaterial];
ENSURES
generatedKey[this, alg];
Above is an excerpt of the rule for SecretKeySpec
. The predicate generatedKey
is listed within the ENSURES
block of this rule. The static analysis labels any object of type SecretKeySpec
by generatedKey
when the analysis finds the object to be used correctly (with respect to its CrySL rule).
On-the-fly Addition or Modification of CrySL Rules
All CrySL rules currently used by CogniCrypt are present in the repository named Crypto-API-Rules. As of April 2020, it contains rules for the four APIs mentioned above. You need to clone the corresponding project and import it as a Maven project into Eclipse where you have already installed CogniCrypt and the CrySL plugins. These plugins let you update the CrySL rules on the fly. You can edit them or even add new rules. CogniCrypt automatically parses these rules and may take them into account in any future analyses and code generations. You need to enable this feature in the CogniCrypt preferences first, though.
The below tutorial describes how to modify CrySL rules on the fly. The first screenshot shows an example code which uses KeyGenerator
that is created with correct algorithm, namely “AES”, and later initialized with a proper keySize i.e. 128. Hence, the plugin doesn’t show any error markers.
Now let us change the keySize
to a incorrect value (Eg. 200) as shown in second screenshot. The plugin displays a error marker upon saving the changes.
The below screenshot shows the value of error marker displayed by the plugin.
The following screenshot shows the original CrySL rule for KeyGenerator
class.
Now let us modify the CrySL rule of KeyGenerator
class so that the init
method also takes 200 as its keySize
and later save the corresponding changes.
Upon saving the new CrySL rule, the plugin would re-run the analysis based your new rules. Consequently, no error markers would be displayed as shown below.
CrySL Rules for the JCA
CogniCrypt ships with a pre-defined set of CrySL rules. The standard rule set covers the correct specification of most classes of the Java Cryptographic Architecture (JCA). The JCA offers various cryptographic services. In the following, we describe these services with their respective classes and briefly summarize important usage constraints. All mentioned classes are defined in the packages javax.crypto
and java.security
of the JCA.
The rule set is also publicly available .The definition of the CrySL rules are found in the files ending in .cryptsl
named with the respective class name.
- Asymmetric Key Generation:
Asymmetric and symmetric cryptography requires different key formats. Asymmetric cryptography uses pairs of public and private keys. While one of the keys encrypts plaintexts to ciphertexts, the second key decrypts the ciphertext. The JCA models a key pair as class
KeyPair
and are generated byKeyPairGenerator
. - Symmetric Key Generation:
Symmetric cryptography uses the same key for encryption and decryption. The JCA models symmetric keys as type
SecretKey
, generated by aSecretKeyFactory
orKeyGenerator
. TheSecretKeyFactory
also enables password-based cryptography usingPBEParameterSpec
orPBEKeySpec
. - Signing and Verification of Data:
The class
Signature
of the JCA allows one to digitally sign data and verify a signature based on a private/public key pair. ASignature
requires the key pair to be correctly generated, hence the rule forSignature
requires a predicate from the asymmetric-key generation task. - Generation of Initialization Vectors:
Initialization vectors (IVs) are used to add entropy to ciphertexts of encryptions. An IV must have enough randomness and must be properly generated. The JCA class
IvParameterSpec
wraps a byte array as an IV and it is required for the array to be randomized bySecureRandom
. The CrySL rule forIvParameterSpec
requires a predicaterandomized
. - Encryption and Decryption
The key component of the JCA is represented by the class
Cipher
, which implements functionality to encrypt or decrypt data. Depending on the used algorithms, modes and paddings must be selected and keys and initialization vectors must be properly generated. Hence, the complete CrySL rule forCipher
requires many other cryptographic services to be executed securely earlier and list them in its respectiveREQUIRES
clause. - Hashing & MACs´:
There are two forms of cryptographic hash functions. A MAC is an authenticated hash that requires a symmetric keys, but there are also keyless hash functions such as MD5 or SHA-256. The JCA’s class
Mac
implements functionality for mac-ing, while keyless hashes are computed byMessageDigest
. - Persisting Keys:
Securely storing key material is an important cryptographic task for confidentiality and integrity of the encrypted data. The JCA class
KeyStore
supports developers in this task and stores the key material. - Cryptographically Secure Random-Number Generation:
Randomness is vital in all aspects of cryptography. Java offers cryptographically secure pseudo-random number generators through
SecureRandom
. As discussed forPBEKeySpec
,SecureRandom
often acts as a helper and therefore many rules list therandomized
predicate in their ownREQUIRES
section.
CrySL Rules for the Bouncy Castle
The below rule set covers the specifications of most classes in the Bouncy Castle (BC). In the following, we describe all the services with their respective classes and briefly summarize important usage constraints. All mentioned classes are defined in the lightweight crypto packages org.bouncycastle.crypto.*
of the BC.
The rule set is also publicly available
- Asymmetric Key Generation:
In BC every asymmetric cryptography has a separate key pair generator. For example RSA has
RSAKeyPairGenerator
, DSA hasDSAKeyPairGenerator
and so on. These asymmetric or public/private, cipher key pair generators should conform to an interfaceAsymmetricCipherKeyPairGenerator
. Every key pair generator has its corresponding key generation parameters which specify the keys being generated. For example, RSA hasRSAKeyGenerationParameters
and DSA hasDSAKeyGenerationParameters
both of which conforms to its base classKeyGenerationParameters
. - Symmetric Key Generation:
The BC has a base class named
CipherKeyGenerator
for symmetric or secret, cipher key generators. Every symmetric algorithm has specific key generator class which extends this base class. For example DES hasDESKeyGenerator
which extends the base class to specify the parameters. - Encryption and Decryption:
There are two variants of
Cipher
equivalent in Bouncy Castle. One isBlockCipher
and the other one isAsymmetricBlockCipher
both of which are interfaces. All the symmetric engines & modes should conform to the former interface and all the asymmetric counterparts should adhere to the latter. BC also provides classes namedBufferedBlockCipher
andBufferedAsymmetricBlockCipher
which are buffer wrappers for block cipher and asymmetric block cipher respectively, allowing the input to be accumulated in a piecemeal fashion until final processing. - Hashing & MACs´:
The BC has a base interface named
Mac
for implementations of message authentication codes (MACs) andDigest
for implementations of hashing. - Cryptographically Secure Random-Number Generation:
The BC uses Java offered cryptographically secure pseudo-random number generator
SecureRandom
for randomness.