- abstract: 'Responsible disclosure is the process by which researchers and vendors cooperate to release information on newly discovered vulnerabilities to the public in an ethically responsible manner. Proper vulnerability disclosure is especially important for the security of embedded firmware in the Internet-of-Things, where a single exploit often impacts thousands of consumer devices. The currently prevalent belief is that disclosing vulnerabilities on some devices is better than not bringing up observed vulnerabilities at all, leaving us with an unknown set of potentially affected devices. Implicitly, this assumes that the potential vulnerability impact of these “invisible” devices is minimal relative to the rest of the publicized set. Yet, should this assumption prove false, a partial reporting of vulnerable devices would conversely pose a greater security risk as malicious actors can trivially use released exploits to target the invisible devices. Our work seeks to quantify the degree to which such vulnerable devices are overlooked during responsible disclosure. We provide a lower-bound estimate of the security impact these “invisible” yet vulnerable devices have for end users. To this end, we model the disclosure process and develop an automated pipeline, BucketLeak, to run a collection of 54 vulnerability exploitation scripts from the years 2010–2025 against a large-scale dataset of 3,569 firmware images belonging to 572 router and camera devices. Our pipeline uncovers 467 unique device-exploit pairs, of which 422 are undisclosed potential N-days that correspond to 290 device models still in circulation. Furthermore, 181 of the models with undisclosed vulnerabilities are still vulnerable even with the latest versions of their firmware installed. By scanning the Internetof-Things with ZoomEye, we find that these 181 vulnerable yet undisclosed devices have more than 1.04 million real-world device counterparts still active and discoverable over the public Internet.',
0 commit comments