diff --git a/mindspore-lite/src/litert/kernel/cpu/nnacl_c/infer/prior_box_infer.c b/mindspore-lite/src/litert/kernel/cpu/nnacl_c/infer/prior_box_infer.c
index a49b2c38e443a1c507892920dc215a742af69cb1..1676c57af028fffff691255c8486f0f442447642 100644
--- a/mindspore-lite/src/litert/kernel/cpu/nnacl_c/infer/prior_box_infer.c
+++ b/mindspore-lite/src/litert/kernel/cpu/nnacl_c/infer/prior_box_infer.c
@@ -69,7 +69,10 @@ int PriorBoxInferShape(const TensorC *const *inputs, size_t inputs_size, TensorC
 
   int32_t min_sizes_size = param->min_sizes_size;
   int32_t max_sizes_size = param->max_sizes_size;
-  int32_t num_priors_box = min_sizes_size * different_aspect_ratios_size + max_sizes_size;
+  NNACL_CHECK_INT_MUL_NOT_OVERFLOW(min_sizes_size, different_aspect_ratios_size, NNACL_ERRCODE_MUL_OVERFLOW);
+  int32_t num_priors_box = min_sizes_size * different_aspect_ratios_size;
+  NNACL_CHECK_INT_ADD_NOT_OVERFLOW(num_priors_box, max_sizes_size, NNACL_ERRCODE_ADD_OVERFLOW);
+  num_priors_box += max_sizes_size;
   const int kPriorBoxPoints = 4;
   const int kPriorBoxN = 1;
   const int kPriorBoxW = 1;
diff --git a/mindspore-lite/src/litert/kernel/cpu/nnacl_c/kernel/group_convolution.c b/mindspore-lite/src/litert/kernel/cpu/nnacl_c/kernel/group_convolution.c
index 9a68f98cda5e6c230c92d6b6748c02a05fe85462..bcf5662bda2c211d18736a7712c4cc20ddacfaa3 100644
--- a/mindspore-lite/src/litert/kernel/cpu/nnacl_c/kernel/group_convolution.c
+++ b/mindspore-lite/src/litert/kernel/cpu/nnacl_c/kernel/group_convolution.c
@@ -304,18 +304,30 @@ int GroupConvolutionCompute(KernelBase *self) {
     // second, separate group conv input into several parts. This step must be in runtime stage.
     int ret = GroupConvSeparateInput(group_conv, i);
     if (ret != NNACL_OK) {
+      self->env_->Free(self->env_->allocator_, sub_kernel_in_tensor->data_);
+      sub_kernel_in_tensor->data_ = NULL;
+      self->env_->Free(self->env_->allocator_, sub_kernel_out_tensor->data_);
+      sub_kernel_out_tensor->data_ = NULL;
       return ret;
     }
 
     // sun kernels run
     ret = group_conv->group_convs_[i]->Compute(group_conv->group_convs_[i]);
     if (ret != NNACL_OK) {
+      self->env_->Free(self->env_->allocator_, sub_kernel_in_tensor->data_);
+      sub_kernel_in_tensor->data_ = NULL;
+      self->env_->Free(self->env_->allocator_, sub_kernel_out_tensor->data_);
+      sub_kernel_out_tensor->data_ = NULL;
       return ret;
     }
 
     // post process, concat all outputs of sub-kernels into one output
     ret = GroupConvPostConcat(group_conv, i);
     if (ret != NNACL_OK) {
+      self->env_->Free(self->env_->allocator_, sub_kernel_in_tensor->data_);
+      sub_kernel_in_tensor->data_ = NULL;
+      self->env_->Free(self->env_->allocator_, sub_kernel_out_tensor->data_);
+      sub_kernel_out_tensor->data_ = NULL;
       return ret;
     }
 
diff --git a/mindspore-lite/src/litert/kernel/cpu/nnacl_c/kernel/scale.c b/mindspore-lite/src/litert/kernel/cpu/nnacl_c/kernel/scale.c
index d87f6dd9e18d117c3b6d6c2ff03a2e94e6b64fd0..cffbb4f858e466bbcb5a69f42e650fc3fcfc34f2 100644
--- a/mindspore-lite/src/litert/kernel/cpu/nnacl_c/kernel/scale.c
+++ b/mindspore-lite/src/litert/kernel/cpu/nnacl_c/kernel/scale.c
@@ -184,7 +184,7 @@ int ScaleInitScaleOffset(ScaleStruct *scale) {
     int malloc_size = NNACLGetElementNum(offset_tensor) * data_type_size;
     NNACL_CHECK_MALLOC_SIZE(malloc_size);
     scale->offset_ = scale->base_.env_->Alloc(scale->base_.env_->allocator_, malloc_size);
-    NNACL_MALLOC_CHECK_NULL_RETURN_ERR(scale->scale_);
+    NNACL_MALLOC_CHECK_NULL_RETURN_ERR(scale->offset_);
     (void)memcpy(scale->offset_, offset_tensor->data_, malloc_size);
   } else {
     scale->malloc_offset_ = false;
diff --git a/mindspore-lite/src/litert/kernel/cpu/nnacl_c/kernel/where.c b/mindspore-lite/src/litert/kernel/cpu/nnacl_c/kernel/where.c
index 15229018f0d7d064b11b4f115c482abbfd58d13b..94a9c3c8075501c78bd9a0149af5544558a9fa6d 100644
--- a/mindspore-lite/src/litert/kernel/cpu/nnacl_c/kernel/where.c
+++ b/mindspore-lite/src/litert/kernel/cpu/nnacl_c/kernel/where.c
@@ -126,6 +126,8 @@ int WhereRunWithSingleInput(WhereStruct *where) {
 
   if (true_num > 0) {
     output->data_ = result;
+  } else {
+    where->base_.env_->Free(where->base_.env_->allocator_, result);
   }
   return NNACL_OK;
 }
diff --git a/mindspore-lite/tools/converter/micro/coder/wrapper/int8/conv1x1_init_int8_wrapper.c b/mindspore-lite/tools/converter/micro/coder/wrapper/int8/conv1x1_init_int8_wrapper.c
index f9ffb0a79d1e7ac264e7542b9d3df2139e170d0e..bac7a523ee7977f466c425668bd1f484ad7c6a27 100644
--- a/mindspore-lite/tools/converter/micro/coder/wrapper/int8/conv1x1_init_int8_wrapper.c
+++ b/mindspore-lite/tools/converter/micro/coder/wrapper/int8/conv1x1_init_int8_wrapper.c
@@ -58,7 +58,6 @@ int Conv1x1Init(int8_t *src_weight, int32_t *src_bias, int32_t *filter_zps, int3
   int32_t *bias_data_ = (int32_t *)(buf + *offset);
   *offset += size * sizeof(int32_t);
   if (bias_data_ == NULL) {
-    free(packed_weight_);
     return NNACL_ERR;
   }
   memset(bias_data_, 0, size * sizeof(int32_t));
@@ -86,7 +85,6 @@ int Conv1x1Init(int8_t *src_weight, int32_t *src_bias, int32_t *filter_zps, int3
   int32_t *bias_data_ = ((*offset + size * sizeof(int32_t)) <= buf_size) ? (int32_t *)(buf + *offset) : NULL;
   *offset += size * sizeof(int32_t);
   if (bias_data_ == NULL) {
-    free(packed_weight_);
     packed_weight_ = NULL;
     return NNACL_ERR;
   }