Create rule S4347: 'SecureRandom' seeds should not be predictable (Part 1) #9300
Conversation
gregory-paidis-sonarsource
force-pushed
the
greg/implement-S4347
branch
2 times, most recently
from
96ddd49
to
d4ef712
Compare
gregory-paidis-sonarsource
added
the
Sprint: MMF-3716 crypto rules
gregory-paidis-sonarsource
force-pushed
the
greg/implement-S4347
branch
6 times, most recently
from
3068250
to
c856b6f
Compare
...src/SonarAnalyzer.CSharp/SymbolicExecution/Roslyn/SecureRandomSeedsShouldNotBePredictable.cs
Show resolved
Hide resolved
...src/SonarAnalyzer.CSharp/SymbolicExecution/Roslyn/SecureRandomSeedsShouldNotBePredictable.cs
Show resolved
Hide resolved
|
|
||
| private static ProgramState ProcessArrayElementReference(ProgramState state, IArrayElementReferenceOperationWrapper arrayElementReference) => | ||
| arrayElementReference.IsAssignmentTarget() || arrayElementReference.IsCompoundAssignmentTarget() | ||
| ? state.SetSymbolConstraint(arrayElementReference.ArrayReference.TrackedSymbol(state), CryptographicSeedConstraint.Unpredictable) |
There was a problem hiding this comment.
Here, if the array access is on the left side of the assignment, I set it to unpredictable.
I could look into the right side of the assignment and see if it has constant value, but I am not sure if there is much value in this case, so I opted for the FN.
...src/SonarAnalyzer.CSharp/SymbolicExecution/Roslyn/SecureRandomSeedsShouldNotBePredictable.cs
Show resolved
Hide resolved
...src/SonarAnalyzer.CSharp/SymbolicExecution/Roslyn/SecureRandomSeedsShouldNotBePredictable.cs
Show resolved
Hide resolved
gregory-paidis-sonarsource
force-pushed
the
greg/implement-S4347
branch
from
c856b6f
to
52683e5
Compare
...src/SonarAnalyzer.CSharp/SymbolicExecution/Roslyn/SecureRandomSeedsShouldNotBePredictable.cs
Show resolved
Hide resolved
| && state[value]?.HasConstraint(CryptographicSeedConstraint.Predictable) is true; | ||
| } | ||
|
|
||
| private static ProgramState ProcessSecureRandomSetSeed(ProgramState state, IInvocationOperationWrapper invocation) |
There was a problem hiding this comment.
This processes SecureRandom.SetSeed(long/byte[])
There was a problem hiding this comment.
The comments you left on the PR helped a lot. Thanks for that.
What do you think about including them in the code?
gregory-paidis-sonarsource
force-pushed
the
greg/implement-S4347
branch
2 times, most recently
from
2a3576c
to
ebe9b96
Compare
...src/SonarAnalyzer.CSharp/SymbolicExecution/Roslyn/SecureRandomSeedsShouldNotBePredictable.cs
Show resolved
Hide resolved
gregory-paidis-sonarsource
force-pushed
the
greg/implement-S4347
branch
2 times, most recently
from
486700e
to
453957b
Compare
|
|
|
@costin-zaharia-sonarsource if you look at the spec:
...it also mentions GUID handling. We could also implement only the following:
What do you think? |
| <p>When using <code>SecureRandom</code>, it is important not to use predictable seeds. This class is used to generate cryptographically strong random | ||
| numbers. Using a predictable seed will make its output predictable as well, which counteracts the use case of <code>SecureRandom</code>.</p> | ||
| <h2>Why is this an issue?</h2> | ||
| <p><code>java.security.SecureRandom</code> is often used to generate random values for cryptographic algorithms. When a random number generator is |
There was a problem hiding this comment.
well, java.security.SecureRandom?
It was probably meant to be Org.BouncyCastle.Security.SecureRandom but you can also check with appsec squad.
There was a problem hiding this comment.
see also the documentation link
There was a problem hiding this comment.
Excellent catch, thanks!
Left a comment on RSPEC PR.
...src/SonarAnalyzer.CSharp/SymbolicExecution/Roslyn/SecureRandomSeedsShouldNotBePredictable.cs
Show resolved
Hide resolved
gregory-paidis-sonarsource
changed the title
github-actions
bot
assigned costin-zaharia-sonarsource and unassigned costin-zaharia-sonarsource
| ? state.SetOperationConstraint(invocation, CryptographicSeedConstraint.Predictable) | ||
| : null; | ||
|
|
||
| private ProgramState ProcessStringToBytes(ProgramState state, IInvocationOperationWrapper invocation) |
There was a problem hiding this comment.
can be static
| && state[value]?.HasConstraint(CryptographicSeedConstraint.Predictable) is true; | ||
| } | ||
|
|
||
| private static ProgramState ProcessSecureRandomSetSeed(ProgramState state, IInvocationOperationWrapper invocation) |
There was a problem hiding this comment.
The comments you left on the PR helped a lot. Thanks for that.
What do you think about including them in the code?
| private static ProgramState ProcessSecureRandomSetSeed(ProgramState state, IInvocationOperationWrapper invocation) | ||
| { | ||
| if (invocation.TargetMethod.Name == "SetSeed" | ||
| && invocation.TargetMethod.ContainingType.DerivesFrom(KnownType.Org_BouncyCastle_Security_SecureRandom) |
There was a problem hiding this comment.
the indentation is off here.
analyzers/tests/SonarAnalyzer.Test/Rules/SecureRandomSeedsShouldNotBePredictableTest.cs
Show resolved
Hide resolved
| @@ -0,0 +1,262 @@ | |||
| using System; | |||
There was a problem hiding this comment.
I think this is very well tested, maybe too well since the symbolic execution engine has it's own tests.
I did not see any tests with inheritance. Would it be worth adding some? SecureRandom is not sealed, although most of the methods we check are static so I'm not sure if it's actually useful.
There was a problem hiding this comment.
I will add a test on part 3, as in part1+2 I do not call new on SecureRandom, only GetInstance, which is a static method that always returns SecureRandom.
github-actions
bot
assigned gregory-paidis-sonarsource and unassigned costin-zaharia-sonarsource
github-actions
bot
moved this from Review in progress
to Review approved
in Best Kanban
|
Since most of the suggestions are not breaking, I will merge this and apply the suggestions on the part#2 PR. |
Part of #8992
Implementing Scenario 1