Python: Make Sanitizer available for urlsplit taint

It isn't used by default, it has to *actively* be enabled.

Author: RasmusWL

python/ql/src/semmle/python/security/strings/External.qll

@@ -195,3 +195,62 @@ class ExternalFileObject extends TaintKind {
         name = "read" and result = this.getValue()
     }
 }
+
+/**
+ * Temporary sanitizer for the tainted result from `urlsplit` and `urlparse`. Can be used to reduce FPs until
+ * we have better support for namedtuples.
+ *
+ * Will clear **all** taint on a test of the kind. That is, on the true edge of any matching test,
+ * all fields/indexes will be cleared of taint.
+ *
+ * Handles:
+ * - `if splitres.netloc == "KNOWN_VALUE"`
+ * - `if splitres[0] == "KNOWN_VALUE"`
+ */
+class UrlsplitUrlparseTempSanitizer extends Sanitizer {
+    // TODO: remove this once we have better support for named tuples
+
+    UrlsplitUrlparseTempSanitizer() { this = "UrlsplitUrlparseTempSanitizer" }
+
+    override predicate sanitizingEdge(TaintKind taint, PyEdgeRefinement test) {
+        (
+            taint instanceof ExternalUrlSplitResult
+            or
+            taint instanceof ExternalUrlParseResult
+        ) and
+        exists(ControlFlowNode foobar |
+            foobar.(SubscriptNode).getObject() = test.getInput().getAUse()
+            or
+            foobar.(AttrNode).getObject() = test.getInput().getAUse()
+        |
+            clears_taint(_, foobar, test.getTest(), test.getSense())
+        )
+    }
+
+    private predicate clears_taint(ControlFlowNode final_test, ControlFlowNode tainted, ControlFlowNode test, boolean sense) {
+        test_equality_with_const(final_test, tainted, sense)
+        or
+        test.(UnaryExprNode).getNode().getOp() instanceof Not and
+        exists(ControlFlowNode nested_test |
+            nested_test = test.(UnaryExprNode).getOperand() and
+            clears_taint(final_test, tainted, nested_test, sense.booleanNot())
+        )
+    }
+
+    /** holds for `== "KNOWN_VALUE"` on `true` edge, and `!= "KNOWN_VALUE"` on `false` edge */
+    private predicate test_equality_with_const(CompareNode cmp, ControlFlowNode operand, boolean sense) {
+        exists(ControlFlowNode const, Cmpop op |
+            const.getNode() instanceof StrConst
+        |
+            (
+                cmp.operands(const, op, operand)
+                or
+                cmp.operands(operand, op, const)
+            ) and (
+                op instanceof Eq and sense = true
+                or
+                op instanceof NotEq and sense = false
+            )
+        )
+    }
+}

python/ql/test/library-tests/taint/namedtuple/Taint.qll

@@ -25,3 +25,23 @@ class DictSource extends TaintSource {
 
     override string toString() { result = "dict taint source" }
 }
+
+class TestConfig extends TaintTracking::Configuration {
+    TestConfig() { this = "TestConfig" }
+
+    override predicate isSanitizer(Sanitizer sanitizer) {
+        sanitizer instanceof UrlsplitUrlparseTempSanitizer
+    }
+
+    override predicate isSource(TaintTracking::Source source) {
+        source instanceof SimpleSource
+        or
+        source instanceof ListSource
+        or
+        source instanceof DictSource
+    }
+
+    override predicate isSink(TaintTracking::Sink sink) {
+        none()
+    }
+}

python/ql/test/library-tests/taint/namedtuple/TestTaint.expected

@@ -3,8 +3,8 @@
 | test.py:13 | test_basic | c | externally controlled string |
 | test.py:13 | test_basic | d | externally controlled string |
 | test.py:13 | test_basic | urlsplit_res | [externally controlled string] |
-| test.py:20 | test_sanitizer | Attribute | externally controlled string |
-| test.py:23 | test_sanitizer | Subscript | externally controlled string |
+| test.py:20 | test_sanitizer | Attribute | NO TAINT |
+| test.py:23 | test_sanitizer | Subscript | NO TAINT |
 | test.py:33 | test_namedtuple | a | NO TAINT |
 | test.py:33 | test_namedtuple | b | NO TAINT |
 | test.py:33 | test_namedtuple | c | NO TAINT |

python/ql/test/library-tests/taint/namedtuple/test.py

@@ -17,10 +17,10 @@ def test_sanitizer():
     urlsplit_res = urlsplit(tainted_string)
 
     if urlsplit_res.netloc == "OK":
-        test(urlsplit_res.netloc) # TODO: FP, should not be tainted here
+        test(urlsplit_res.netloc)
 
     if urlsplit_res[2] == "OK":
-        test(urlsplit_res[0]) # TODO: FP, should not be tainted here
+        test(urlsplit_res[0])
 
 def test_namedtuple():
     tainted_string = TAINTED_STRING