Code Security
Gordon College
Stephen Brinton
Virtual Machine Security
• Building a fence around your code
– JVM – Java Virtual Machine
• Originally developed by Sun Microsystems
• Executes Java Bytecode
– Execution by: Interpretation (JVM) or Compilation (JIT)
– Common Language Runtime
• Developed by Microsoft
• Executes Common Intermediate Language (CIL)
– Verified and then run as native code on machine
Java’s Security
• Strongly Typed Compiler
– Eliminate programming bugs
– Help enforce language semantics
• Bytecode verifier
– Makes sure the rules of Java are followed in the compiled
code
• Classloader
– Finds, loads, and defines classes (runs verifier on them)
• Security Manager
– Main interface between the system and the Java Code
Trusted vs Untrusted Code
• Trusted
– Java API code
– Code loaded from the classpath
– resides outside the “sandbox”
• Untrusted
– Code loaded from outside the classpath (usually from a
network)
– Confined to the “sandbox”
• Java Apps (by default) live outside the sandbox and
Java Applets are confined within the sandbox
What’s a “Sandbox”?
• Applets run inside a “sandbox”
– If you download code, it has to play within
the JVM (Sandbox)
– SecurityManager is
called for certain
methods, and can
forbid access
• JDK1.1 introduced the notions of codesigning and “trusted applets”
Sandboxes
How do you protect your computer bad
code?
• The solution: Make untrusted code play within
a sandbox.
• Need for varying security policies increases
– You assign “permissions” to pieces of code
– JDK 1.1 (digital signatures) – if user trusts the
digitally signed code – users could allow normally
untrusted code to access resources
• Enforcement Mechanism for Policy
(sandbox)?
– Security Managers
Java Sandbox
The sandbox for untrusted Java applets, for
example, prohibits many activities, including:
• Reading or writing to the local disk
• Making a network connection to any host, except
the host from which the applet came
• Creating a new process
• Loading a new dynamic library and directly calling
a native method
Java Sandbox
• The fundamental components responsible for
Java's sandbox are:
* Safety features built into the Java virtual
machine (and the language)
* The class loader architecture
* The class file verifier
* The security manager and the Java API
Creating a security manager in JDK 1.1
(that allows reading files, but disallows writing files)
public class MySecurityManager extends
java.lang.SecurityManager {
public void checkRead(String file) throws SecurityException {
// reading is allowed, so just return
return;
}
public void checkWrite(String file) throws SecurityException {
// writing is not allowed, so throw the exception
throw new SecurityException("Writing is not allowed");
}
} // end MySecurityManager
Creating a security manager in JDK 1.1
(that allows reading files, but only with extension “txt”)
public class MySecurityManager2 extends
java.lang.SecurityManager {
public void checkRead(String file) throws SecurityException {
//check the file extension to see if it ends in ".txt"
int index=file.lastIndexOf('.');
String result=file.substring(index, file.length());
if(result.equalsIgnoreCase(".txt")){
return;
}else{
throw new SecurityException("Cannot read file: "+file);
}
}
}
Security Manager prior to JDK 1.2
Only way security was controlled prior to JDK 1.2
Advantages:
– Easy to provide a binary security model (yes, you can or no, you can't)
– Methods in the SecurityManager class are called by the Java
API; there's no need for you to call the code at all
– Interface of this system is constant across all JVM platforms;
one security manager can run everywhere
– Additional size of a simple security manager is negligible
– Security manager & class loader work hand-in-hand to
ensure neither is compromised by accident or an act of evil
Security Manager prior to JDK 1.2
Disadvantages:
– Control of security is in the developer's hands, not
in a security specialist's hands
– Not easy to provide a customizable security model
that varies from user to user
– Only way to change an existing policy is to change
or subclass the existing security manager; not all
users have the capability of programming in Java
– New security policies (non-system resource
policies, for example) are difficult to implement
Install a SecurityManager
• Applications don’t start, by
default, with a SecurityManager
• You must install one, either
from within the code, or
using a command line
argument
java -Djava.security.manager
Building a bigger and better sandbox
• To provide greater control to user and
developer – Traditional sandbox (java.security
package) was expanded to include:
• AccessController class
– Muscle of the security manager – enforces policy
• Permission class
• Policy class
AccessController
java -new -usepolicy FileApp test.txt
The Policytool
Policy Tool Main Window
The Policytool
Policy Tool Edit Entry Window
The Policytool
Policy Tool Add Grant Entry Window
.NET Framework
• Different administrative software model
• Fixed location (traditional)
• Dynamic nature of software (present)
– Dynamic downloads and execution
– Remote execution
– Security is essential
Security Models
• Role-based security
– Users have access to resources based on roles
– Model used by most operating systems
• Code access security (new with .Net)
– Also called “evidence-based security”
– Even if user is trusted - the code may not be.
– Tackles the problem with mobile code
Both models are found in .Net framework
Code Access Security
mechanism of the CLR
– Manages code and depends on level of trust
– CLR – will lessen and tighten its grip based on
permission and trust level
• Very similar to the sandbox view
– Two aspects:
• Control the access level given to an application
(assembly)
• Control access to a particular resource (like a database)
.Net execution
• Runtime framework
– Runs both managed and unmanaged code
• Managed - under control of runtime
– Has access to certain features: memory management, JIT,
and security services
– MSIL (MS intermediate language)
– Can be compiled to native code prior to execution
• Unmanaged - compiled for a certain system
– Can not directly use the runtime
MSIL : object-oriented assembly language for
an abstract, stack-based machine
CIL
common intermediate language
C#
public double GetVolume()
{
double volume =
height*width*thickness;
if(volume<0) return 0;
return volume;
}
object-oriented
assembly
language for an
abstract, stackbased machine
.method public hidebysig instance float64
GetVolume() cil managed
{
// Code size 51 (0x33)
.maxstack 2
.locals init ([0] float64 volume,
[1] float64 CS$00000003$00000000)
IL_0000: ldarg.0
IL_0001: ldfld float64 OOP.Aperture::height
IL_0006: ldarg.0
IL_0007: ldfld float64 OOP.Aperture::width
IL_000c: mul
IL_000d: ldarg.0
IL_000e: ldfld float64
OOP.Aperture::thickness
IL_0013: mul
IL_0014: stloc.0
IL_0015: ldloc.0
IL_0016: ldc.r8 0.0
IL_001f: bge.un.s IL_002d
IL_0021: ldc.r8 0.0
IL_002a: stloc.1
IL_002b: br.s IL_0031
IL_002d: ldloc.0
IL_002e: stloc.1
IL_002f: br.s IL_0031
IL_0031: ldloc.1
IL_0032: ret
} // end of method Aperture::GetVolume
CLR
common language runtime
CLR
common language runtime
3 different runtimes
1. A single CLR that runs all ASP.NET apps
2. Browsers (IE) uses a single CLR that
executes all dowloaded controls
3. CLR used to execute commands run from
the OS shell.
Verification
• 2 Forms of verification done at runtime
– MSIL - verified
• Invalid - JIT compiler cannot make native exe
• Valid - can be made into native code
• Type safe - interacts with types through
exposed contracts
• Verifiable - can be proved to be type-safe
– Assembly metadata - validated
• Metadata - describes aspects of the code file
Verification
• Integrated with the compiler
• If code is trusted
skip verification and compile
otherwise
MSIL verification
assembly metadata verification
if successful verification - compile
Code Access Security
• Assigns permissions to assemblies
based on assembly evidence
• Evidence identifies code: location, etc.
• Evidence is attached to assembly when
loaded for execution
• Evidence limits what program can do
Opt-in approach
Permissions
• Authority to perform protected
operations
– Accessing files, registry, network, GUI,
execution environment, skip verification
• Assembly load time
– Evidence -> grant permissions
Evidence
•
•
•
•
•
•
•
Zone
URL
Hash (encrypted value)
Strong name - unique ID for program
Site
Application Directory
Publisher certificate
Evidence
<System.Security.Policy.Zone version="1">
<Zone>MyComputer</Zone>
</System.Security.Policy.Zone>
<System.Security.Policy.Url version="1">
<Url>
file:///C:/winnt/microsoft.net/framework/v1.0.2728/mscorlib.dll
</Url>
</System.Security.Policy.Url>
<StrongName version="1"
Key="00000000000000000400000000000000"
Name="mscorlib"
Version="1.0.2411.0"/>
<System.Security.Policy.Hash version="1">
<RawData>4D5A90000300000004000000FFFF0000B8000000000000...
0000000000000000000000000000000000000000000000000000
</RawData>
</System.Security.Policy.Hash>
Evidence Based Security
• The loader discovers evidence of the origin of the
code
– Evidence is info about the assembly
– Used to determine the permissions granted to an assembly
– Evidence is the input to the security policy
•
•
•
•
•
Publisher
Strong Name
Site
URL
Zone
Authenticode signer
public key+name+version
Web site of code origin
URL of code origin
zone of code origin
– Extensible for new kinds of evidence (custom)
Security Policy
Mapping
{Assembly
Evidence}
Control by administrator
Policy levels:
* Enterprise Policy Level
* Machine Policy Level
* User Policy Level
* Application Domain Policy Level
{Permissions
Granted}
Security Policy
Determining the set of granted permissions:
1. Policy levels evaluate the evidence and
generate a set of permissions.
2. Permission sets calculated for each policy
level are intersected with each other.
3. Resulting permission set is compared with
the set of permissions the assembly declared
necessary to run.
Security Policy
.NET Framework
• Code groups
– Bring together code with similar characteristics
• Evidence
– Information used to place code into code groups
– where code is from (internet or intranet), publisher, strong name,
URI from download, etc.
• Arranged in a hierarchy
• Permissions
– Actions you allow each code group to perform
• For example: “able to access the user interface”
– Managed by system admin. at the enterprise, machine and
user levels
.NET Framework
.NET Framework
Condition: All code, Permission Set: Nothing
Condition: Zone: Internet, Permission Set: Internet
Condition: URL: www.monash.edu.au, Permission Set: MonashPSet
Condition: Strong Name: m-Commerce, Permission Set: m-CommercePSet
Microsoft Management Console snap-in
Condition: All code, Permission Set: Nothing
Condition: Zone: Internet, Permission Set: Internet
Condition: URL: www.monash.edu.au, Permission Set: MonashPSet
Condition: Strong Name: m-Commerce, Permission Set: m-CommercePSet
Permission Sets
•
•
•
•
FullTrust: Allows unrestricted access to system resources.
SkipVerification: Allows an assembly to skip verification.
Execution: Allows code to execute.
Nothing: No permissions. Not granting the permission to
execute effectively stops code from running
• Internet: Appropriate for code coming from the Internet. (Limited)
Code will not receive access to the file system or registry, but
can do limited user interface actions as well as use the safe file
system called Isolated Storage.
Predefined Psets - can be accessed within code.
Stack Walk
• When are the permissions generated by the
code-access security module checked?
– To determine whether code is authorized to access a
resource or perform an operation, the runtime's security
system walks the call stack, comparing the granted
permissions of each caller to the permission being
demanded.
a protected resource may demand a stack walk
Stack Walk
• When are the permissions generated by the
code-access security module checked?
– To determine whether code is authorized to access a
resource or perform an operation, the runtime's security
system walks the call stack, comparing the granted
permissions of each caller to the permission being
demanded.
Security Policy Topology
• Multiple policy levels of administration
– Enterprise: policy distributed across organization
– Machine: policy for all users of a machine
– User: policy specific to logged on user
• Effective policy is the intersection of levels
Enterprise policy
Machine1 policy
User
A
User
B
Machine2 policy
User
C
User
D
Evaluating Policy Per Level
• Each Policy Level contains a set of code groups
– Code groups are arranged in a tree
– Every code group has a membership condition and a set of
granted permissions
– An assembly is mapped to one of more code groups based
on the evidence that the assembly provides
YES
NO
Internet?
ALL CODE? None
YES
P1
NO
Intranet?
P2
Local?
P3
Completing Policy Resolve
• Matching code groups evaluated
– Union of matching permission sets per level, this
is the permissions allowed by this level
– Intersection of Policy Levels produces the final
ALLOWED permission set for the assembly
• ALLOWED = Enterprise
Machine
User
Policy Tools
• caspol.exe
• Console Management snap-in
– .NET Framework 2.0 Configuration
Example
caspol.exe –ld
look at the code groups on a machine
caspol.exe –listgroups
look at the code groups on a machine (more compact)
caspol.exe –resolvegroup simpleSecure.exe
view an assembly’s code groupings (effective permission – intersection)
caspol.exe –resolveperm simpleSecure.exe
view an assembly’s permissions (each code groups brings additional
permissions)
.NET Framework 2.0 Configuration
Detail Information
“Building a bigger sandbox”
http://www.javaworld.com/javaworld/jw-08-1998/jw-08-sandboxp2.html
“Security in the .NET Framework”
http://msdn2.microsoft.com/en-us/library/fkytk30f(vs.80).aspx
“Java vs. .NET Security”
http://www.onjava.com/pub/a/onjava/2003/11/26/javavsdotnet.html