---
title: "驯服安全与能耗悖论：面向 Android 恶意软件检测的绿色 AI 优化方法"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.20003v1"
permalink: "https://feed.astralor.com/items/itm_895bbe575b1dd843b0daab05"
published: "2026-07-22T10:39:42.000Z"
score: 74
channel: "tech"
tags: ["Android恶意软件检测","arXiv","Green AI","INT8量化","MLP"]
---

# 驯服安全与能耗悖论：面向 Android 恶意软件检测的绿色 AI 优化方法

## AI 点评

> 这篇论文最值得关注的是它用最朴素的全连接网络（MLP）加上量化，就解决了看似矛盾的安全与能耗问题，说明很多时候不需要复杂模型。对做移动端AI部署的工程师来说，文中具体的能耗数据（0.0189 mJ/推理）和层数选择建议可以直接复用。

## 速览

该研究针对 Android 设备上深度学习模型带来的高能耗问题，提出通过 INT8 量化与浅层 MLP 架构平衡恶意软件检测性能与能效。实验表明，INT8 量化使模型体积缩小约 3.5 倍，每次推理能耗降至 0.0189 mJ，同时保持 99.2% 以上的检测准确率。3 层和 4 层量化神经网络通过提升吞吐量、缩短 CPU 高功率运行时间进一步降低能耗，为资源受限智能手机上的绿色安全检测提供了可行方案。

## 关键点

- INT8 量化使模型体积缩小约 3.5 倍，每次推理能耗降至 0.0189 mJ
- 浅层量化架构（3 层和 4 层 MLP）在保持 99.2% 以上准确率的同时显著降低能耗
- 通过提升吞吐量、缩短 CPU 高功率运行时间实现节能

## 全文

An increase in advanced Android malware requires the use of deep learning models, which can run on Android devices. But there is a trade-off between security and energy use, as strong detection models can drain the battery of devices fast. This work tests different Multi-Layer Perceptron (MLP) model configurations to balance malware detection performance and energy efficiency. In this work, we compared standard FP32 models with optimized INT8 quantized neural networks with different model depths using TUANDROMD and DREBIN datasets for both classification performance and energy consumption. The results show that INT8 quantization reduces model size by about 3.5 times with a decrease in energy consumption to 0.0189 mJ per inference, while maintaining more than 99.2\% detection accuracy. We found that shallow quantized architectures, such as 3-layer and 4-layer QNNs, reduce energy costs by improving throughput and shortening the time of CPU operating in a high-power state. This work shows that efficient malware protection can be achieved on resource-constrained smartphones and provides a foundation for Green AI in mobile security.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_895bbe575b1dd843b0daab05