OWASP Orizon Project XML

The Orizon check XML schema

Security checks can be divided in:

• design_check
• keyword_check
• execution_check
• stats_check

Statistics check

The stats check XML schema has been changed from version 1.0, so this applies starting from version 1.1 and later

<stats
    subj=[code | comment | complexity]
    verb=[lt | gt | le | ge | ne | eq | ratio]
    [ direct_object= [loc | loC] ]
    [modifier = "percentage"]
    value=numeric value
  />

where:

• Subject can be one of the following:
  • code: line of code
  • comment: line of comments
  • complexity: ciclomatic complexity index

• verb is the boolean comparison operator between the subject and the value:
  • lt: lesser than
  • gt: grater than
  • le: lesser or equal than
  • ge: greater or equal than
  • ne: not equal than
  • eq: equal than
  • ratio: indicates the ratio subj versus direct_object

Design check

 This is a work in progress section. Actually I'm collecting all the security checks coming from other tools and trying to figure it out the XML schema to use to 
 describe security checks.
 thesp0nge, 07.12.2008

Classes

 <design
       subj="class"
       name=ignored
       verb=[counts|contains|extends|implements|has]
       direct_object=[scope|attribute]
       value=the value being checked
 >
 </design>

The class design will be evaluated for the number in the source file, the classes it extends or implements and for modifiers such as scope and attribute. With the counts verb, it will be evaluated how many classes are contained into a specific source file. For a non object oriented program, the class it will be ignored and inteded to be the whole program.

 Example.
 
 If we need to have our J2EE applications comprised of source files containing 5 classes each, this is the correspondent security check
 <design subj="class" verb="counts" value="5" />

The extends verb must be used if a security check needs to evaluate the class extended by the one being evaluated.

 Example
 
 If a security check needs to evaluate if a class extends org.owasp.orizon.About, the check can be written using the following notation
 <design subj="class" verb="extends" value="org.owasp.orizon.About" />

The implements verb is the same as the previous except that the check is intended to be applied for interfaces or abstract classes the one being reviewed implements.

The has verb is the only one that makes meaningful the direct_object attribute with this rationale:

• scope: the parameter to be evaluated is the class scope

 Example
 
 If in your organization you must check that all Java classes has to be public for such a reason we don't know to care about, you can write the security check this way
 <design subj="class" verb="has" direct_object="scope" value="public" />

• attribute: an attribute can be if the class is inner or not

keyword_check, about keyword specific checks

  <keyword
    name=keyword name
  />

execution_check: extra care must be taken for parameter in this desing...

  <exec
    caller_class=a class name
    caller_method=a method name
  />

The Orizon Input file XML schema

Orizon 1.0 will bring 3 new subsystems in Jericho engine:

• local analisys (control flow graph)
• global analisys (call graph)
• taint propagation analisys (data graph)

Each of this subsystems will use a different input file provided by the translator, so each source file will be translated in 3 different XML files with different schema of course.

Taint propagation analisys

This subsystem is devoted to analyze variable content and how data is managed by the application.

Here is the schema to be used to describe a generic operation over a variable or a socket or a generic I/O operation.

 <taint
      subj="[variable|socket|sql|file]"
      name="the variable name"
      type="the variable data type"
      verb="[created|modified|deleted|read_data|write_data]"
      constant="[yes|no]"
      must_reduce="[yes|no]"
      value"the value being used to fill the variable"
 >
      expression to be reduced...
 </taint>

Here there are some example to understand better how instructions over variable, will be translated to XML.

• • Variable declaration

To better describe variable declaration we must discriminate from simple variable rather than complex objects in OO programming languages.

If we need to declare a brand new integer value,

  int a;

we will obtain

  <taint subj="variable" name="a" type="int" verb="created" constant="" value="" must_reduce="no" />

If we choose to create a new variable with a init value,

  int b = 3;

we will obtain

  <taint subj="variable" name="b" type="int" verb="created" constant="" value="3" must_reduce="no" />

Now lets create an object rather than a simple variable.

  String c = new String("A new string");

The correspondent XML will be

  <taint subj="variable" name="c" type="String" verb="created" constant="" value="A new string" must_reduce="no" />

What happens if some complex constructor call is issued

  String d = new String((new Integer(3)).toString());

This example is quite odd, but it's a common practice to have very complex object constructor calls.

  <taint subj="variable" name="d" type="String" verb="created" constant="" value="" must_reduce="yes">
      <taint subj="variable" name="dyno1" type="Integer" verb="created" value="3" must_reduce="no">
      <call variable="dyno1" method="toString()">
            <result type="String">3</result>
      </call>  
  </taint>

• • Generic non constant assignment

Let a be an integer variable. We want to stuff volatile value of '5' into it...

  a = 5;

it becomes

  <taint subj="variable" name="a" verb="modified" constant="no" value="5" />