diff --git a/dnn/src/aarch64/matrix_mul/algos.cpp b/dnn/src/aarch64/matrix_mul/algos.cpp
index 0b640d3289d7115040d8099d4905df9f70434fa4..13104c8cd942af9788640effe173197a1f8b9ed3 100644
--- a/dnn/src/aarch64/matrix_mul/algos.cpp
+++ b/dnn/src/aarch64/matrix_mul/algos.cpp
@@ -908,6 +908,9 @@ bool MatrixMulImpl::AlgoInt8x8x16MK4_16x12x4::preferred(
     return false;
 #else
     auto arch = cpuinfo_get_current_core()->uarch;
+#ifdef __IN_TEE_ENV__
+    arch = cpuinfo_uarch_unknown;
+#endif
     bool little_core = arch == cpuinfo_uarch_cortex_a53 ||
                        arch == cpuinfo_uarch_cortex_a55;
     return little_core;
@@ -986,6 +989,9 @@ bool MatrixMulImpl::AlgoInt8x8x16MK4_4x4x8::preferred(
     return false;
 #else
     auto arch = cpuinfo_get_current_core()->uarch;
+#ifdef __IN_TEE_ENV__
+    arch = cpuinfo_uarch_unknown;
+#endif
     bool little_core = arch == cpuinfo_uarch_cortex_a53 ||
                        arch == cpuinfo_uarch_cortex_a55;
     return !little_core;
diff --git a/dnn/src/aarch64/matrix_mul/fp32/strategy.cpp b/dnn/src/aarch64/matrix_mul/fp32/strategy.cpp
index 14c1238bda2c3d837d458ba21451afe564fcb7f8..9b5d513847d98ce62bca0f85d96388c65a50981d 100644
--- a/dnn/src/aarch64/matrix_mul/fp32/strategy.cpp
+++ b/dnn/src/aarch64/matrix_mul/fp32/strategy.cpp
@@ -181,6 +181,9 @@ void sgemm_8x12::kern(const float* packA, const float* packB, size_t M,
                                            is_first_k);
 #else
     auto arch = cpuinfo_get_current_core()->uarch;
+#ifdef __IN_TEE_ENV__
+    arch = cpuinfo_uarch_unknown;
+#endif
     if (arch == cpuinfo_uarch_cortex_a53) {
         sgemm_8x12_helper<matmul_general_8x12_a53>(packA, packB, M, N, K, C,
                                                    LDC, is_first_k);
@@ -274,6 +277,9 @@ void sgemm_mk4_8x12::kern(const float* packA, const float* packB, size_t M,
                                            is_first_k);
 #else
     auto arch = cpuinfo_get_current_core()->uarch;
+#ifdef __IN_TEE_ENV__
+    arch = cpuinfo_uarch_unknown;
+#endif
     if (arch == cpuinfo_uarch_cortex_a53) {
         sgemm_mk4_8x12_helper<matmul_mk4_8x12_a53>(packA, packB, M, N, K, C,
                                                    LDC, is_first_k);
diff --git a/dnn/src/arm_common/conv_bias/fp32/direct_kernels/f32_direct_nchw_nchw44_kern_common.h b/dnn/src/arm_common/conv_bias/fp32/direct_kernels/f32_direct_nchw_nchw44_kern_common.h
index 9184f278c3f40ee6e186458be249de564415e5b3..8090c5e5ce2c9005f365fc59cc1e299d45be1324 100644
--- a/dnn/src/arm_common/conv_bias/fp32/direct_kernels/f32_direct_nchw_nchw44_kern_common.h
+++ b/dnn/src/arm_common/conv_bias/fp32/direct_kernels/f32_direct_nchw_nchw44_kern_common.h
@@ -851,6 +851,10 @@ struct ConvDirectFp32NchwNchw44<bias_mode, Op, 3, 2> {
 #else
         auto arch_tag = CpuTag::A7_TAG;
 #endif
+
+#ifdef __IN_TEE_ENV__
+        arch_tag = CpuTag::DEFAULT_CPU_TAG;
+#endif
         if (arch_tag == CpuTag::A7_TAG) {
             for (int oc_idx = 0; oc_idx < oc_end; oc_idx += big_oc_step) {
                 const int weight_offset = oc_idx * ic * fh * fw;