樂鑫為上海證券交易所科創板的上市公司，ESP32是該公司的主要產品，合作對象從百度、小米到Amazon不等，2019年首次公開發行股票（IPO）時的牌價為62.6元人民幣，上周的收盤價是231.5元人民幣。

這些隱藏命令屬於主機控制器命令（Host Controller Interface，HCI），總計有29個，包括寫入記憶體在內，已被賦予漏洞編號CVE-2025-27840，可允許駭客修改晶片以解鎖其它功能，植入惡意程式，或是進行裝置身分竊盜攻擊。

意謂著駭客可冒充已知裝置連結到手機、電腦或其它智慧裝置，以獲取存在於這些裝置上的機密資訊或商業對話，也可作為監視之用。

Tarlogic是利用該公司所開發的安全稽核工具BluetoothUSB，找到了ESP32的隱藏功能，這是一個免費的跨平臺驅動程式，可用來執行安全測試與模擬攻擊。Tarlogic表示，BluetoothUSB允許用戶不必使用不同的硬體即可完成所有測試，降低測試門檻。

樂鑫的英文論壇已出現針對此事的討論，當中有一名自稱是樂鑫員工的ESP_Sprite表示，儘管他只看到簡報，沒看到詳細資訊，但他認為大多數的問題可以透過韌體更新來解決，少數的問題則需要搭配其它安全漏洞才能被利用，且迄今尚未看到針對相關漏洞的概念性驗證。

新聞來源：iThome

Since its establishment in 2005, Winstec Corp. has been committed to the development and integration of embedded real-time systems and wireless smart solutions. In the past, it has designed and manufactured many devices and accumulated rich experience as well as a powerful tool library. As we take a look at the past development cases of Winstec, the many cases vary from car navigation, medical imaging, and smart access control intercom system to driving recorder, smart nursing brush, and Bluetooth intercom headset. Although these products seem to be extremely different, they are in fact well-aligned with a clear technical context as the Winstec team specializes in the design of embedded systems, image processing, network technology, electrical control systems, and wireless smart solutions. Based on these foundations, R&D into innovative applications and new technology were then conducted according to the actual needs of different projects.

Among the projects executed by Winstec, about 70% are based on customer designs or ODM requirements, and the other 30% are based on independent R&D of innovative solutions. For the company, the goal is to quickly provide innovative integrated applications to customers and launch them to the market. Winstec's project development experiences in AI imaging, such as object recognition, thermal imaging lens, and AI face recognition machine, happen to coincide with the EVB needs which the IoT Service Hub wishes to develop. Thus, a device equipped with image recognition function and AI computing capability—the "HUB 8735" EVB—was materialized with the input of joint efforts.

This EVB integrates the most advanced NPU AI computing engine, allowing HUB 8735 to efficiently process AI models. At the same time, HUB 8735 is also compatible with Arduino development features, and has a variety of pre-trained AI models that are already optimized on the module. Such features are expected to greatly simplify the work of developers and allow them to deploy AI applications more efficiently. As for the software and hardware design of the EVB, there is external development instructions synced with the ESP32-CAM module to allow seamless integration during development. It can be used by simply replacing the module. With all the other popular modules on the market, the planning and designing of "MIT" EVBs will require substantial efforts and careful layout.

Among many AIoT products, the efficient image processing capability of HUB 8735 and its multi-functional communication methods allow it to play a key role in various scenarios, from smart homes to industrial IoT, from smart retail to healthcare and automotive electronics, covering many aspects of everyday life.