adding IDA extractor code to resolve nested data references for string and bytes features

capa/features/extractors/ida/helpers.py

@@ -331,3 +331,28 @@ def is_basic_block_tight_loop(bb):
             if ref == bb.start_ea:
                 return True
     return False
+
+
+def find_data_reference_from_insn_helper(ea, max_depth=10):
+    """ recursive search for data reference, if exists, from instruction by resolving nested pointers, if exist """
+    if 0 == max_depth:
+        # return when max depth reached
+        return ea
+
+    data_refs = list(idautils.DataRefsFrom(ea))
+
+    if len(data_refs) != 1:
+        # return if no refs or more than one ref (assume nested pointers only have one data reference)
+        return ea
+
+    if ea == data_refs[0]:
+        # return if circular reference
+        return ea
+
+    # continue searching
+    return find_data_reference_from_insn_helper(data_refs[0], max_depth - 1)
+
+
+def find_data_reference_from_insn(insn):
+    """ return address of data reference, if exists, otherwise address of instruction """
+    return find_data_reference_from_insn_helper(insn.ea)

capa/features/extractors/ida/insn.py

@@ -116,11 +116,8 @@ def extract_insn_bytes_features(f, bb, insn):
         example:
             push    offset iid_004118d4_IShellLinkA ; riid
     """
-    if idaapi.is_call_insn(insn):
-        # ignore call instructions
-        return
-
-    for ref in idautils.DataRefsFrom(insn.ea):
+    ref = capa.features.extractors.ida.helpers.find_data_reference_from_insn(insn)
+    if ref != insn.ea:
         extracted_bytes = capa.features.extractors.ida.helpers.read_bytes_at(ref, MAX_BYTES_FEATURE_SIZE)
         if extracted_bytes and not capa.features.extractors.helpers.all_zeros(extracted_bytes):
             yield Bytes(extracted_bytes), insn.ea
@@ -137,7 +134,8 @@ def extract_insn_string_features(f, bb, insn):
         example:
             push offset aAcr     ; "ACR  > "
     """
-    for ref in idautils.DataRefsFrom(insn.ea):
+    ref = capa.features.extractors.ida.helpers.find_data_reference_from_insn(insn)
+    if ref != insn.ea:
         found = capa.features.extractors.ida.helpers.find_string_at(ref)
         if found:
             yield String(found), insn.ea

tests/test_ida_features.py

@@ -100,6 +100,13 @@ def test_string_features():
     assert capa.features.String("bcrypt.dll") not in features
 
 
+@pytest.mark.skip(reason="IDA Pro tests must be run within IDA")
+def test_string_pointer_features():
+    f = get_extractor().get_function(0x0044EDEF)
+    features = extract_function_features(f)
+    assert capa.features.String("INPUTEVENT") in features
+
+
 @pytest.mark.skip(reason="IDA Pro tests must be run within IDA")
 def test_byte_features():
     f = get_extractor().get_function(0x40105D)
@@ -109,6 +116,13 @@ def test_byte_features():
     assert wanted.evaluate(features) == True
 
 
+@pytest.mark.skip(reason="IDA Pro tests must be run within IDA")
+def test_bytes_pointer_features():
+    f = get_extractor().get_function(0x0044EDEF)
+    features = extract_function_features(f)
+    assert capa.features.Bytes("INPUTEVENT".encode("utf-16le")).evaluate(features) == True
+
+
 @pytest.mark.skip(reason="IDA Pro tests must be run within IDA")
 def test_number_features():
     f = get_extractor().get_function(0x40105D)