基础块条数精简

exe2json.py

@@ -14,7 +14,7 @@ import multiprocessing
 
 ret_trap_opcode_family = ["ret", "hlt", "int3", "ud2"]
 
-sample_type = 'malware'
+sample_type = 'benign'
 
 
 def extract_opcode(disasm_text):
@@ -71,6 +71,7 @@ def get_graph_cfg_r2pipe(r2pipe_open, file_path, output_path, file_name):
                 block_addr = block["addr"]
                 block_Statement = []
 
+
                 node_list.append(block["addr"])
                 # 获取基本块的反汇编指令
                 disasm = r2pipe_open.cmdj("pdj " + str(block["ninstr"]) + " @" + str(block["addr"]))
@@ -78,17 +79,13 @@ def get_graph_cfg_r2pipe(r2pipe_open, file_path, output_path, file_name):
                 # if len(disasm) != 49144:
                 #     continue
                 if disasm:
-                    last_op = ''
-                    # if len(disasm) > 200:
-                    #     logger.warning(
-                    #         f"基础块指令长度异常，{file_path}，函数名称{function['name']}基础块地址{block['addr']},长度{len(disasm)}")
+                    block_len = len(disasm)
                     for op_index, op in enumerate(disasm):
-                        op_disasm = extract_opcode(op["disasm"])
-                        # 防止大量重复的语句造成内存溢出
-                        if op_disasm == last_op:
-                            continue
-                        last_op = op_disasm
                         # 提取操作码并转换为bert模型输入格式
+                        op_disasm = extract_opcode(op["disasm"])
+                        # 如果单个基础块的长度大于20且操作码重复，则跳过
+                        if block_len > 20 and op_disasm in block_Statement:
+                            continue
                         block_Statement.append(op_disasm)
                         # 处理跳转码并构建cfg
                         if 'jump' in op:
@@ -129,6 +126,8 @@ def get_graph_cfg_r2pipe(r2pipe_open, file_path, output_path, file_name):
                                 # 当前指令是基础块的最后一条指令
                                 if op not in ret_trap_opcode_family or op["type"] not in ["ret", "trap"]:
                                     temp_edge_list.append([block_addr, op["offset"] + op["size"]])
+                    if block_len > 20:
+                        logger.warning(f"二进制可执行文件解析警告，基础块长度大于20，文件{file_path},基础块地址{block_addr}，操作码长度{block_len}->{len(block_Statement)}")
 
                 # debugger
                 # print(len(disasm))
@@ -257,7 +256,7 @@ if __name__ == '__main__':
     sample_file_list = os.listdir(sample_file_path)
     print(f"max worker {os.cpu_count()}")
     with multiprocessing.Pool(processes=os.cpu_count()) as pool:
-        result = list(tqdm(pool.imap_unordered(exe_to_json, [os.path.join(sample_file_path, file_name) for file_name in sample_file_list]), total=len(sample_file_list)))
+        result = list(tqdm(pool.imap_unordered(exe_to_json, [os.path.join(sample_file_path, file_name) for file_name in sample_file_list[::-1]]), total=len(sample_file_list)))
 
 
     # with concurrent.futures.ThreadPoolExecutor(max_workers=os.cpu_count()) as executor: