* Added slice bounds testing for slice expressions.
* Added checking slice index.
* Added test for reassigning slice.
* Store capacities on reslicing.
* Scope change clears map. Func name used to track slices.
* Map CallExpr to check bounds when passing to functions.
* Fixed linter errors.
* Updated rulelist with CWE mapping.
* Added comment for NewSliceBoundCheck.
* Addressed nil cap runtime error.
* Replaced usage of nil in call arg map with dummy callexprs.
* Updated comments, wrapped error return, addressed other review concerns.
* G101 now checks LHS of ValueAssignments for patternValue.
* Added matching string literals in equality check.
* Added patternValue matching for ValueSpec.
* Ran gci to fix linter error.
* Added tests and updated regex to be more inclusive.
* Addressed short-circuit eval for isHighEntropy and non-standard ok variable.
* Resolved unhandled error and added more tests.
* Flattened code to make it more readable.
* Added better comments.
* Added new regex for Google API Key, GitHub PAT, and GoogleOAuth.
* Gofmt'ed the test cases.
* Remove read only types from unsafe defer rules
* Remove rule G307 which checks when an error is not handled when a file or socket connection is closed
This doesn't seem to bring much value from security perspective, and it caused a lot of controversy since
is a very common pattern in Go.
* Mentioned in documentation that rule G307 is retired
* Clean up the test for rule G307
The existing code assumed imports to be either imported, or imported with an
alias. Badly formatted files may have duplicate imports for a package, using
different aliases.
This patch refactors the code, and;
Introduces a new `GetImportedNames` function, which returns all name(s) and
aliase(s) for a package, which effectively combines `GetAliasedName` and
`GetImportedName`, but adding support for duplicate imports.
The old `GetAliasedName` and `GetImportedName` functions have been rewritten to
use the new function and marked deprecated, but could be removed if there are no
external consumers.
With this patch, the linter is able to detect issues in files such as;
package main
import (
crand "crypto/rand"
"math/big"
"math/rand"
rand2 "math/rand"
rand3 "math/rand"
)
func main() {
_, _ = crand.Int(crand.Reader, big.NewInt(int64(2))) // good
_ = rand.Intn(2) // bad
_ = rand2.Intn(2) // bad
_ = rand3.Intn(2) // bad
}
Before this patch, only a single issue would be detected:
gosec --quiet .
[main.go:14] - G404 (CWE-338): Use of weak random number generator (math/rand instead of crypto/rand) (Confidence: MEDIUM, Severity: HIGH)
13:
> 14: _ = rand.Intn(2) // bad
15: _ = rand2.Intn(2) // bad
With this patch, all issues are identified:
gosec --quiet .
[main.go:16] - G404 (CWE-338): Use of weak random number generator (math/rand instead of crypto/rand) (Confidence: MEDIUM, Severity: HIGH)
15: _ = rand2.Intn(2) // bad
> 16: _ = rand3.Intn(2) // bad
17: }
[main.go:15] - G404 (CWE-338): Use of weak random number generator (math/rand instead of crypto/rand) (Confidence: MEDIUM, Severity: HIGH)
14: _ = rand.Intn(2) // bad
> 15: _ = rand2.Intn(2) // bad
16: _ = rand3.Intn(2) // bad
[main.go:14] - G404 (CWE-338): Use of weak random number generator (math/rand instead of crypto/rand) (Confidence: MEDIUM, Severity: HIGH)
13:
> 14: _ = rand.Intn(2) // bad
15: _ = rand2.Intn(2) // bad
While working on this change, I noticed that ImportTracker.TrackFile() was not able
to find import aliases; Analyser.Check() called both ImportTracker.TrackFile() and
ast.Walk(), which (with the updated ImportTracker) resulted in importes to be in-
correctly included multiple times (once with the correct alias, once with the default).
I updated ImportTracker.TrackFile() to fix this, but with the updated ImportTracker,
Analyser.Check() no longer has to call ImportTracker.TrackFile() separately, as ast.Walk()
already handles the file, and will find all imports.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
It appears that `GetImportedName` returns _both_ aliased and non-aliased imports.
As a result, a file having both crypto/rand and math/rand (but aliased) would
trigger false positives on G404. Given the following file;
```go
package main
import (
"crypto/rand"
"math/big"
rnd "math/rand"
)
func main() {
_, _ = rand.Int(rand.Reader, big.NewInt(int64(2)))
_ = rnd.Intn(2)
}
```
And patching for debugging;
```patch
diff --git a/helpers.go b/helpers.go
index 437d032..80f4233 100644
--- a/helpers.go
+++ b/helpers.go
@@ -250,6 +250,8 @@ func GetBinaryExprOperands(be *ast.BinaryExpr) []ast.Node {
// GetImportedName returns the name used for the package within the
// code. It will ignore initialization only imports.
func GetImportedName(path string, ctx *Context) (string, bool) {
+ fmt.Printf("%+v", ctx.Imports.Imported)
+ os.Exit(1)
importName, imported := ctx.Imports.Imported[path]
if !imported {
return "", false
```
Would show that `math/rand` was included in the list, using it's non-aliased
name (`:rand`).
gosec -quiet .
map[crypto/rand:rand math/big:big math/rand:rand]
This patch works around this problem by reversing the order in which imports
are resolved in `MatchCallByPackage()`. Aliased packages are tried first, after
which non-aliased imports are tried.
Given the example application mentioned above:
Before this patch:
```bash
gosec -quiet .
Results:
[/Users/sebastiaan/Projects/test/gosec-issue/main.go:10] - G404 (CWE-338): Use of weak random number generator (math/rand instead of crypto/rand) (Confidence: MEDIUM, Severity: HIGH)
9: func main() {
> 10: _, _ = rand.Int(rand.Reader, big.NewInt(int64(2)))
11: _ = rnd.Intn(2)
```
With this patch applied:
```bash
gosec --quiet .
Results:
[/Users/sebastiaan/Projects/test/gosec-issue/main.go:11] - G404 (CWE-338): Use of weak random number generator (math/rand instead of crypto/rand) (Confidence: MEDIUM, Severity: HIGH)
10: _, _ = rand.Int(rand.Reader, big.NewInt(int64(2)))
> 11: _ = rnd.Intn(2)
12: }
```
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
* Add check for usage of Rat.SetString in math/big with an overflow error
Rat.SetString in math/big in Go before 1.16.14 and 1.17.x before 1.17.7
has an overflow that can lead to Uncontrolled Memory Consumption.
It is the CVE-2022-23772.
* Use ContainsPkgCallExpr instead of manual parsing
* Find G303 in string concatenations, with os.TempDir, and in path.Join args
* Find G303 with /usr/tmp, too
/usr/tmp is commonly found e.g. on Solaris.
In Go 1.16 or higher, the `io/ioutil` has been deprecated and the
`ioutil.ReadFile` function now calls `os.ReadFile`.
Signed-off-by: Eng Zer Jun <engzerjun@gmail.com>
* Add G307 sample code.
The sample should reflect a defered close that leads to data loss.
Due to IDE auto-complete people tend at least log errors, but not
really care about handling.
* Add more G307 sample code. Propose a way to implement
* Remove unused code. Add example that should not return an error but does
* Remove test for synced closed file for now.
Will add this later
Co-authored-by: Cosmin Cojocar <cosmin.cojocar@gmx.ch>
* fix: add variable assignment checking as part of MinVersion
* fix: add more code to allow assignment with const
* fix: rework the code and add more test cases for MinVersion
* fix: format linting issue using gofumpt
