merge upstream

Co-authored-by: Willi Ballenthin <wballenthin@google.com>

CHANGELOG.md

@@ -83,6 +83,7 @@ Special thanks to our repeat and new contributors:
 - binja: fix and simplify stack string detection code after binja 4.0 @xusheng6
 - binja: add support for forwarded export #1646 @xusheng6
 - cape: support more report formats #2035 @mr-tz
+- elf: extract import / export symbols from stripped binaries #2096 @ygasparis
 
 ### capa explorer IDA Pro plugin
 - replace deprecated IDA API find_binary with bin_search #1606 @s-ff

capa/features/extractors/elffile.py

@@ -10,8 +10,7 @@ import logging
 from typing import Tuple, Iterator
 from pathlib import Path
 
-from elftools.elf.elffile import ELFFile, SymbolTableSection
-from elftools.elf.relocation import RelocationSection
+from elftools.elf.elffile import ELFFile, DynamicSegment, SymbolTableSection
 
 import capa.features.extractors.common
 from capa.features.file import Export, Import, Section
@@ -47,17 +46,37 @@ def extract_file_export_names(elf: ELFFile, **kwargs):
 
             yield Export(symbol.name), AbsoluteVirtualAddress(symbol.entry.st_value)
 
+    for segment in elf.iter_segments():
+        if not isinstance(segment, DynamicSegment):
+            continue
+
+        logger.debug("Dynamic Segment contains %s symbols: ", segment.num_symbols())
+
+        for symbol in segment.iter_symbols():
+            # The following conditions are based on the following article
+            # http://www.m4b.io/elf/export/binary/analysis/2015/05/25/what-is-an-elf-export.html
+            if not symbol.name:
+                continue
+            if symbol.entry.st_info.type not in ["STT_FUNC", "STT_OBJECT", "STT_IFUNC"]:
+                continue
+            if symbol.entry.st_value == 0:
+                continue
+            if symbol.entry.st_shndx == "SHN_UNDEF":
+                continue
+
+            yield Export(symbol.name), AbsoluteVirtualAddress(symbol.entry.st_value)
+
 
 def extract_file_import_names(elf: ELFFile, **kwargs):
     # Create a dictionary to store symbol names by their index
     symbol_names = {}
 
     # Extract symbol names and store them in the dictionary
-    for section in elf.iter_sections():
-        if not isinstance(section, SymbolTableSection):
+    for segment in elf.iter_segments():
+        if not isinstance(segment, DynamicSegment):
             continue
 
-        for _, symbol in enumerate(section.iter_symbols()):
+        for _, symbol in enumerate(segment.iter_symbols()):
             # The following conditions are based on the following article
             # http://www.m4b.io/elf/export/binary/analysis/2015/05/25/what-is-an-elf-export.html
             if not symbol.name:
@@ -73,21 +92,19 @@ def extract_file_import_names(elf: ELFFile, **kwargs):
 
             symbol_names[_] = symbol.name
 
-    for section in elf.iter_sections():
-        if not isinstance(section, RelocationSection):
+    for segment in elf.iter_segments():
+        if not isinstance(segment, DynamicSegment):
             continue
 
-        if section["sh_entsize"] == 0:
-            logger.debug("Symbol table '%s' has a sh_entsize of zero!", section.name)
-            continue
+        relocation_tables = segment.get_relocation_tables()
+        logger.debug("Dynamic Segment contains %s relocation tables:", len(relocation_tables))
 
-        logger.debug("Symbol table '%s' contains %s entries:", section.name, section.num_relocations())
-
-        for relocation in section.iter_relocations():
-            # Extract the symbol name from the symbol table using the symbol index in the relocation
-            if relocation["r_info_sym"] not in symbol_names:
-                continue
-            yield Import(symbol_names[relocation["r_info_sym"]]), FileOffsetAddress(relocation["r_offset"])
+        for relocation_table in relocation_tables.values():
+            for relocation in relocation_table.iter_relocations():
+                # Extract the symbol name from the symbol table using the symbol index in the relocation
+                if relocation["r_info_sym"] not in symbol_names:
+                    continue
+                yield Import(symbol_names[relocation["r_info_sym"]]), FileOffsetAddress(relocation["r_offset"])
 
 
 def extract_file_section_names(elf: ELFFile, **kwargs):

tests/test_elffile_features.py

@@ -14,17 +14,11 @@ from capa.features.extractors.elffile import extract_file_export_names, extract_
 
 CD = Path(__file__).resolve().parent
 SAMPLE_PATH = CD / "data" / "055da8e6ccfe5a9380231ea04b850e18.elf_"
+STRIPPED_SAMPLE_PATH = CD / "data" / "bb38149ff4b5c95722b83f24ca27a42b.elf_"
 
 
-def test_elffile_import_features():
-    expected_imports = [
-        "memfrob",
-        "puts",
-        "__libc_start_main",
-        "malloc",
-        "__cxa_finalize",
-    ]
-    path = Path(SAMPLE_PATH)
+def check_import_features(sample_path, expected_imports):
+    path = Path(sample_path)
     elf = ELFFile(io.BytesIO(path.read_bytes()))
     # Extract imports
     imports = list(extract_file_import_names(elf))
@@ -40,6 +34,52 @@ def test_elffile_import_features():
         assert symbol_name in extracted_symbol_names, f"Symbol '{symbol_name}' not found in imports."
 
 
+def check_export_features(sample_path, expected_exports):
+    path = Path(sample_path)
+    elf = ELFFile(io.BytesIO(path.read_bytes()))
+    # Extract imports
+    exports = list(extract_file_export_names(elf))
+
+    # Verify that at least one export was found
+    assert len(exports) > 0, "No exports were found."
+
+    # Extract the symbol names from the extracted imports
+    extracted_symbol_names = [exported[0].value for exported in exports]
+
+    # Check if all expected symbol names are found
+    for symbol_name in expected_exports:
+        assert symbol_name in extracted_symbol_names, f"Symbol '{symbol_name}' not found in exports."
+
+
+def test_stripped_elffile_import_features():
+    expected_imports = ["__cxa_atexit", "__cxa_finalize", "__stack_chk_fail", "fclose", "fopen", "__android_log_print"]
+    check_import_features(STRIPPED_SAMPLE_PATH, expected_imports)
+
+
+def test_stripped_elffile_export_features():
+    expected_exports = [
+        "_ZN7_JNIEnv14GetArrayLengthEP7_jarray",
+        "Java_o_ac_a",
+        "Java_o_ac_b",
+        "_Z6existsPKc",
+        "_ZN7_JNIEnv17GetStringUTFCharsEP8_jstringPh",
+        "_ZN7_JNIEnv21GetObjectArrayElementEP13_jobjectArrayi",
+        "_ZN7_JNIEnv21ReleaseStringUTFCharsEP8_jstringPKc",
+    ]
+    check_export_features(STRIPPED_SAMPLE_PATH, expected_exports)
+
+
+def test_elffile_import_features():
+    expected_imports = [
+        "memfrob",
+        "puts",
+        "__libc_start_main",
+        "malloc",
+        "__cxa_finalize",
+    ]
+    check_import_features(SAMPLE_PATH, expected_imports)
+
+
 def test_elffile_export_features():
     expected_exports = [
         "deregister_tm_clones",
@@ -55,17 +95,4 @@ def test_elffile_export_features():
         "_IO_stdin_used",
         "__libc_csu_init",
     ]
-    path = Path(SAMPLE_PATH)
-    elf = ELFFile(io.BytesIO(path.read_bytes()))
-    # Extract imports
-    exports = list(extract_file_export_names(elf))
-
-    # Verify that at least one export was found
-    assert len(exports) > 0, "No exports were found."
-
-    # Extract the symbol names from the extracted imports
-    extracted_symbol_names = [exported[0].value for exported in exports]
-
-    # Check if all expected symbol names are found
-    for symbol_name in expected_exports:
-        assert symbol_name in extracted_symbol_names, f"Symbol '{symbol_name}' not found in exports."
+    check_export_features(SAMPLE_PATH, expected_exports)