diff --git a/include/api/cell.h b/include/api/cell.h
index e0813467666..fa41d061921 100644
--- a/include/api/cell.h
+++ b/include/api/cell.h
@@ -84,7 +84,7 @@ class MS_API InputAndOutput {
  private:
   std::shared_ptr<CellBase> cell_;
   std::vector<InputAndOutput> prev_;
-  int32_t index_;
+  int32_t index_ = 0;
 };
 }  // namespace mindspore
 #endif  // MINDSPORE_INCLUDE_API_CELL_H
diff --git a/include/api/types.h b/include/api/types.h
index fa5e1d53732..9a1d6dd6291 100644
--- a/include/api/types.h
+++ b/include/api/types.h
@@ -352,8 +352,8 @@ void MSTensor::SetTensorName(const std::string &name) { return SetTensorName(Str
 
 using Key = struct Key {
   const size_t max_key_len = 32;
-  size_t len;
-  unsigned char key[32];
+  size_t len = 0;
+  unsigned char key[32] = {0};
   Key() : len(0) {}
   explicit Key(const char *dec_key, size_t key_len);
 };
diff --git a/mindspore/lite/src/common/decrypt.cc b/mindspore/lite/src/common/decrypt.cc
index d7be33c8894..23a56960965 100644
--- a/mindspore/lite/src/common/decrypt.cc
+++ b/mindspore/lite/src/common/decrypt.cc
@@ -74,7 +74,7 @@ bool ParseEncryptData(const Byte *encrypt_data, size_t encrypt_len, std::vector<
   }
   int_buf.assign(encrypt_data, encrypt_data + sizeof(int32_t));
   auto iv_len = ByteToInt(int_buf.data(), int_buf.size());
-  if (iv_len <= 0 || iv_len + sizeof(int32_t) + sizeof(int32_t) > encrypt_len) {
+  if (iv_len <= 0 || static_cast<size_t>(iv_len) + sizeof(int32_t) + sizeof(int32_t) > encrypt_len) {
     MS_LOG(ERROR) << "assign len is invalid.";
     return false;
   }
@@ -87,7 +87,8 @@ bool ParseEncryptData(const Byte *encrypt_data, size_t encrypt_len, std::vector<
   }
 
   (*iv).assign(encrypt_data + sizeof(int32_t), encrypt_data + sizeof(int32_t) + iv_len);
-  if (cipher_len <= 0 || sizeof(int32_t) + iv_len + sizeof(int32_t) + cipher_len > encrypt_len) {
+  if (cipher_len <= 0 ||
+      sizeof(int32_t) + static_cast<size_t>(iv_len) + sizeof(int32_t) + static_cast<size_t>(cipher_len) > encrypt_len) {
     MS_LOG(ERROR) << "assign len is invalid.";
     return false;
   }
@@ -300,7 +301,7 @@ std::unique_ptr<Byte[]> Decrypt(const std::string &lib_path, size_t *decrypt_len
       MS_LOG(ERROR) << "The block_size read from the cipher data must be not negative, but got " << block_size;
       return nullptr;
     }
-    if (offset + block_size > data_size) {
+    if (offset + static_cast<size_t>(block_size) > data_size) {
       MS_LOG(ERROR) << "assign len is invalid.";
       return nullptr;
     }
diff --git a/mindspore/lite/src/common/quant_utils.cc b/mindspore/lite/src/common/quant_utils.cc
index d0ef3ccbdcd..757c3c80c8f 100644
--- a/mindspore/lite/src/common/quant_utils.cc
+++ b/mindspore/lite/src/common/quant_utils.cc
@@ -103,7 +103,7 @@ int CalQuantizationParams(schema::QuantParamT *quant_param, double real_min, dou
 int GetBucketIndex(const std::vector<int> &dims, int preferred_dim, int data_index) {
   int stride = 1;
   int bucket_count = dims[preferred_dim];
-  for (size_t i = preferred_dim + 1; i < dims.size(); i++) {
+  for (size_t i = static_cast<size_t>(preferred_dim + 1); i < dims.size(); i++) {
     stride *= dims[i];
   }
   if (stride == 0 || bucket_count == 0) {
@@ -134,8 +134,11 @@ int CalPerChannelGain(size_t bit_num, const std::vector<int> &dims, int preferre
   const int bits_per_byte = 8;
   const int quant_param_size = 32;
   int channels = dims.at(preferred_dim);
-  CHECK_LESS_RETURN(channels, 1);
-  size_t bucket_size = elem_count / channels;
+  if (channels < 1) {
+    MS_LOG(ERROR) << "channels must not less 1";
+    return RET_ERROR;
+  }
+  size_t bucket_size = static_cast<size_t>(elem_count / channels);
   bool do_quant = (quant_param_size * bits_per_byte) / (sizeof(float) * bits_per_byte - bit_num) < bucket_size;
   if (do_quant) {
     return RET_OK;
@@ -157,7 +160,7 @@ int CalWeightQuantBias(const float *raw_datas, size_t elem_count, const std::vec
   std::map<int, double> var_raws;
   std::map<int, double> var_dequants;
   size_t bucket_size = quant_params->size();
-  int bucket_volume = elem_count / dims[preferred_dim];
+  int bucket_volume = static_cast<size_t>(elem_count / dims[preferred_dim]);
   // Init Map
   for (size_t i = 0; i < bucket_size; i++) {
     total_raws[i] = 0;
diff --git a/mindspore/lite/src/common/string_util.cc b/mindspore/lite/src/common/string_util.cc
index 4d0ce31c070..41687114ba7 100644
--- a/mindspore/lite/src/common/string_util.cc
+++ b/mindspore/lite/src/common/string_util.cc
@@ -59,7 +59,8 @@ int WriteStringsToTensor(Tensor *tensor, const std::vector<StringPack> &string_b
   }
   size_t num = string_buffer.size();
   std::vector<int32_t> offset(num + 1);
-  offset[0] = 4 * (num + 2);
+  const size_t extra_offset_num = 2;
+  offset[0] = static_cast<int32_t>(sizeof(int32_t) * (num + extra_offset_num));
   for (size_t i = 0; i < num; i++) {
     offset[i + 1] = offset[i] + string_buffer[i].len;
   }
@@ -92,7 +93,8 @@ int WriteSeperatedStringsToTensor(Tensor *tensor, const std::vector<std::vector<
   }
   size_t num = string_buffer.size();
   std::vector<int32_t> offset(num + 1);
-  offset[0] = 4 * (num + 2);
+  const size_t extra_offset_num = 2;
+  offset[0] = static_cast<int32_t>(sizeof(int32_t) * (num + extra_offset_num));
   std::vector<int> len(num);
   for (size_t i = 0; i < num; i++) {
     len[i] = 0;