Connectedness: A Dimension of Security Bug Severity Assessment For Measuring Uncertainty
Chan Shue Long · 2025
Read Connectedness: A Dimension of Security Bug Severity Assessment For Measuring Uncertainty by Chan Shue Long (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-17813.txt.
Introduction
Shue Long Chan argues that security bug scores miss a whole dimension: how connected a bug is. The standard scoring system weighs how easy a bug is to exploit against the harm it does. That covers the known facts. It says little about the fog around a bug, where many exploit paths and side effects may lurk.
Chan proposes connectedness, a count of how strongly a bug ties to other entities in a system. Wide ties mean wider possible impact and more doubt about the true score. The short paper makes the case for adding this axis to severity calls.
Catalog record
| CREATOR | Chan Shue Long |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 14 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-17813
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Connectedness: A Dimension of Security Bug Severity Assessment For Measuring Uncertainty Chan Shue Long Abstract Current frameworks for evaluating security bug severity, such as the Common Vulnerability Scoring System (CVSS), prioritize the ratio of exploitability to impact. This paper suggests that the above approach measures the "known knowns" but inadequately addresses the "known unknowns" especially when there exist multiple possible exploit paths and side effects, which introduce significant uncertainty. This paper introduces the concept of connectedness, which measures how strongly a security bug is connected with different entities, thereby reflecting the uncertainty of impact and the exploit potential. This work highlights the critical but underappreciated role connectedness plays in severity assessments. Keywords Attack Surface \cdot Cyber Risk Quantification \cdot Philosophy of Cybersecurity \cdot Uncertainty 1 Introduction The adage "we can not control what we can not measure." ( Verendel 2009 ) underscores the centrality of quantification in cybersecurity. Yet, as Knight and Jones ( Knight and Jones 2002 ) observed, the conflation of risk and uncertainty persists 1 1 1 The practical difference between risk and uncertainty is ”that in the former, the distribution of the outcome in a group of instances is known (either through calculation a priori or from statistics of past experience), while in the case of uncertainty this is not true” ( Knight and Jones 2002 ) . We can hereby define risk as the ”known knowns” and uncertainty as the ”unknown unknowns” as suggested by Simpson ( Simpson 2024 ) . For example, a microservice having a 3 percent chance of melting down and being out of service is a risk; knowing that a feature of a given application is likely to be problematic without knowing its exact problems and how likely they occur, is an uncertainty. . For instance, CVSS’s exploitability and impact metrics collapse multi-faceted vulnerabilities into singular values, disregarding the uncertainty inherent in unexplored attack vectors and side effects. Today’s severity assessment frameworks usually prioritize two dimensions: (1) the likelihood of a vulnerability being exploited and (2) the severity of its impact. The state-of-the-art scoring system CVSS measures them correspondingly with "exploitability metrics" and "impact metrics" ( Mell et al. 2022 ) 2 2 2 also see https://www.first.org/cvss/v4.0/specification-document ), and they are only capable of capturing the risk of a single exploit path. When there exist multiple possible exploit paths and multiple possible side effects, CVSS fails to address this situation and forces the user to use a single value (for each sub-item in the metrics) to represent the exploitability of multiple possible exploit paths and the impact of multiple possible side effects - in such cases, it fails to see a security bug as a set of possible security events and measure the effect of the induced uncertainty. In such situations, uncertainty plays an important role in severity analysis; hence, we need a way to measure the impact of uncertainty. We will argue that connectedness is a good measure of uncertainty. First, we need to establish the view: "security bug is a set of possible security events". 2 Security bug is a set of possible security events A security bug can have multiple possible exploit paths and multiple possible side effects. Let us consider XSS as an example. In earlier days when XSS prevention techniques were not mature and feasible libraries were not abundant, people used custom filters to sanitize user input before reflecting it in the response, such as catching and removing dangerous elements in the input, HTML-encode the input, etc. However, these methods were not very effective, due to their bypasses of which many offensive security practitioners must be familiar with. The main cause of the difficulties of the early XSS defense attempts faced is that there were many possible exploit paths, even if they were not known until someone published them: the user input reflected in the response body has connections with many different entities such as: 1. an HTML element 2. an attribute of an existing element 3 3 3 such as the href attribute of an existing a tag 3. the inside of existing JavaScript code 4. how HTML is parsed in different browsers and different libraries 4 4 4 for example see https://portswigger.net/research/bypassing-dompurify-again-with-mutation-xss If we consider the impact of an XSS, we will also see multiple possible side effects, such as cookie stealing 5 5 5 which is usually the main concern. , phishing 6 6 6 by disguising a phishing attempt as a genuine content served by a trusted domain, through its url. , cross-origin cookie leakage 7 7 7 in case the cookie of the another site is scoped to parent domain, and the vulnerable site is another subdomain. This is hypothetical, but still a possible side effect, although not a strong one. . Hence, one should not see XSS as a single event, but as a set of possible events including: 1. the user input creates a dangerous HTML entity 2. the user input creates an entity with a dangerous attribute 3. the user input creates a dangerous attribute in an existing entity 4. the user input creates an entity which confuses the HTML parsing logic 3 Connectedness 3.1 The more uncertainty, the more potential risk With hindsight, even if we lived in the early 2000s, we should have expected many possible ways to exploit XSS utilizing different objects related to the logic of how a webpage is rendered and how HTML is parsed. For example, even if we do not know that it is possible to exploit it by utilizing event handlers when the input is reflected in the href attribute of a <a> tag, we should still see it as a possible way. We do not know the actual way to exploit, but we do know that the behavior "user input getting reflected in the response" is connected with all components related to how a webpage is served, in particular the <a> tag and its href attribute; we just do not know exactly how and how likely it allows an exploit - this is the "known unknown". Once we find out an actual exploit, it becomes a "known known". The more "known unknowns", the more uncertainty we have, and the more we should expect that some of them will one day turn into "known knowns". Equivalently, if a risk is a potential negative security event, we can see "uncertainty" as a potential risk. 3.2 Definition of connectedness We hereby introduce the concept of connectedness : given a behavior (a system’s behavior of interest: including security bugs), connectedness is a quantity describing the intensity of the behavior with other entities. The intensity consists of the number of entities connected with the given behavior and the strength of their connection. The more and stronger connections a given behavior has, the higher its connectedness. The strength of a connection is determined by how closely the given behavior is related to the entity. In the case of XSS, the behavior of interest is that the user input pollutes the response and causes unexpected JavaScript execution. The entities connected with this behavior consist of HTML elements, attributes of existing elements, the inside of existing JavaScript code and the parser differentials 2 . The first three