Fix byte/string extraction and unit tests (#1339)

* Fix wrong expected results on string and bytes tests. Fix https://github.com/mandiant/capa/issues/1336

* Fix IDA insn/byte extractor checks wrong address. Fix https://github.com/mandiant/capa/issues/1327

* fix vivisect string check and tests

---------

Co-authored-by: Xusheng <xusheng@vector35.com>

CHANGELOG.md

@@ -25,6 +25,7 @@
 ### Bug Fixes
 - extractor: fix vivisect loop detection corner case #1310 @mr-tz
 - match: extend OS characteristic to match OS_ANY to all supported OSes #1324 @mike-hunhoff
+- extractor: fix IDA and vivisect string and bytes features overlap and tests #1327 #1336 @xusheng6 
 
 ### capa explorer IDA Pro plugin
 

capa/features/extractors/ida/insn.py

@@ -172,7 +172,7 @@ def extract_insn_bytes_features(fh: FunctionHandle, bbh: BBHandle, ih: InsnHandl
     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):
-            if not capa.features.extractors.ida.helpers.find_string_at(insn.ea):
+            if not capa.features.extractors.ida.helpers.find_string_at(ref):
                 # don't extract byte features for obvious strings
                 yield Bytes(extracted_bytes), ih.address
 

capa/features/extractors/viv/insn.py

@@ -175,8 +175,13 @@ def derefs(vw, p):
     while True:
         if not vw.isValidPointer(p):
             return
+
         yield p
 
+        if vw.isProbablyString(p) or vw.isProbablyUnicode(p):
+            # don't deref strings that coincidentally are pointers
+            return
+
         try:
             next = vw.readMemoryPtr(p)
         except Exception:
@@ -271,7 +276,7 @@ def extract_insn_bytes_features(fh: FunctionHandle, bb, ih: InsnHandle) -> Itera
             if capa.features.extractors.helpers.all_zeros(buf):
                 continue
 
-            if f.vw.isProbablyString(v):
+            if f.vw.isProbablyString(v) or f.vw.isProbablyUnicode(v):
                 # don't extract byte features for obvious strings
                 continue
 

tests/fixtures.py

@@ -337,6 +337,9 @@ def get_sample_md5_by_name(name):
         return "946a99f36a46d335dec080d9a4371940"
     elif name.startswith("b9f5b"):
         return "b9f5bd514485fb06da39beff051b9fdc"
+    elif name.startswith("294b8d"):
+        # file name is SHA256 hash
+        return "3db3e55b16a7b1b1afb970d5e77c5d98"
     else:
         raise ValueError("unexpected sample fixture: %s" % name)
 
@@ -643,14 +646,19 @@ FEATURE_PRESENCE_TESTS = sorted(
         # insn/string, direct memory reference
         ("mimikatz", "function=0x46D6CE", capa.features.common.String("(null)"), True),
         # insn/bytes
-        ("mimikatz", "function=0x40105D", capa.features.common.Bytes("SCardControl".encode("utf-16le")), True),
+        ("mimikatz", "function=0x401517", capa.features.common.Bytes(binascii.unhexlify("CA3B0E000000F8AF47")), True),
-        ("mimikatz", "function=0x40105D", capa.features.common.Bytes("SCardTransmit".encode("utf-16le")), True),
+        ("mimikatz", "function=0x404414", capa.features.common.Bytes(binascii.unhexlify("0180000040EA4700")), True),
-        ("mimikatz", "function=0x40105D", capa.features.common.Bytes("ACR  > ".encode("utf-16le")), True),
+        # don't extract byte features for obvious strings
+        ("mimikatz", "function=0x40105D", capa.features.common.Bytes("SCardControl".encode("utf-16le")), False),
+        ("mimikatz", "function=0x40105D", capa.features.common.Bytes("SCardTransmit".encode("utf-16le")), False),
+        ("mimikatz", "function=0x40105D", capa.features.common.Bytes("ACR  > ".encode("utf-16le")), False),
         ("mimikatz", "function=0x40105D", capa.features.common.Bytes("nope".encode("ascii")), False),
+        # push    offset aAcsAcr1220 ; "ACS..." -> where ACS == 41 00 43 00 == valid pointer to middle of instruction
+        ("mimikatz", "function=0x401000", capa.features.common.Bytes(binascii.unhexlify("FDFF59F647")), False),
         # IDA features included byte sequences read from invalid memory, fixed in #409
         ("mimikatz", "function=0x44570F", capa.features.common.Bytes(binascii.unhexlify("FF" * 256)), False),
-        # insn/bytes, pointer to bytes
+        # insn/bytes, pointer to string bytes
-        ("mimikatz", "function=0x44EDEF", capa.features.common.Bytes("INPUTEVENT".encode("utf-16le")), True),
+        ("mimikatz", "function=0x44EDEF", capa.features.common.Bytes("INPUTEVENT".encode("utf-16le")), False),
         # insn/characteristic(nzxor)
         ("mimikatz", "function=0x410DFC", capa.features.common.Characteristic("nzxor"), True),
         ("mimikatz", "function=0x40105D", capa.features.common.Characteristic("nzxor"), False),