Security Analysis of Chain-FS service August 9, 2026

Vanessa Teague, Arash Mirzaei · 2025

Read Security Analysis of Chain-FS service August 9, 2026 by Vanessa Teague, Arash Mirzaei (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-2503-23627.txt.

Introduction

Vanessa Teague and Arash Mirzaei audit the security of a cloud storage service that advertises itself as trustless, in a 2025 paper. Files stored with the service, called Chain-FS, are encrypted end to end, which should keep even the operator from reading them. The authors examined how that encryption is actually built and how sharing works.

They find two weaknesses. First, the way passwords feed the encryption allows a dictionary attack by a malicious server whenever a password only just meets the minimum length rule. Weak-but-allowed passwords can then be guessed offline at scale. Second, the sharing feature sends the decryption passwords to the server protected only by the standard transport encryption, handing the operator the very keys the design claims never to hold. The paper undercuts the trustless claim with concrete, checkable flaws.

Catalog record

CREATORVanessa Teague, Arash Mirzaei
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE14 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-23627

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Security Analysis of Chain-FS service August 9, 2026 Vanessa Teague, Arash Mirzaei Abstract We examine the security of a cloud storage service that makes very strong claims about the “trustless” nature of its security. We find that, although stored files are end-to-end encrypted, the encryption method allows for effective dictionary attacks by a malicious server when passwords only just meet the minimum length required. Furthermore, the file sharing function simply sends the decryption passwords to the server with no protection other than TLS. I Introduction Chain-FS ( https://www.chain-fs.com/ ) is a file storage and sharing system that claims very strong privacy properties. This is the way Chain-FS describes itself: “Chain-FS is a secure storage API that allows for the immutable storage of data, as well as providing the data owner with complete control of information sharing, and unprecedented privacy. Chain-FS is a trust-less architecture, meaning that unlike traditional cloud stores, Chain-FS delivers a platform that says, there is no need to trust us with your data, because we cannot read it, share, or modify it, and this includes our system administrators.’’ 1 1 1 https://www.chain-fs.com/about Last accessed Mar 31, 2025. These sorts of claims are made prominently: a screenshot of the main landing page, with similar assurances, is shown in Figure 1 . Fig. 1: The main landing page of chain-fs.com, as at 31 Mar 2025. Claims that even administrators cannot read the data feature prominently. The purpose of this work is to analyze the security features of Chain-FS and investigate if it is able to provide a trust-less architecture. We analyzed the JavaScript code executed in the browser, which was accessible through the browser’s developer tools. We identified two main problems. 1. File encryption can be easily reversed through systematic password-guessing, when passwords meeting the minimal accepted length are used. We were able to iterate through all permitted 6-character passwords on an ordinary laptop in around 30 hours. This means that an attacker with control of the server can read at least some properly-encrypted files. The details and suggested corrections are in Section II . Since our initial report, the minimum length has increased to 8 characters, which makes no substantial difference. 2. File sharing simply sends the passwords to the server in the clear (apart from TLS). This undermines the security goals of an end-to-end encrypted service and means that the architecture relies entirely on trusting the server not to decrypt a file that has been shared. It is also vulnerable to any attackers able to subvert the TLS connection, including proxies. The details are in Section III . Section II describes the implementation of encryption including some technical errors. Section IV gives an overview of other issues and questions. Section V describes the disclosure history of this report. These shortcomings are particularly important given the way this product is marketed. Chain-FS is actively marketed in Australia to survivors of domestic violence. Since Australia has laws for compelling data access from corporations, the promise of a trustless architecture, with the technical guarantees of end-to-end encryption, may be particularly critical for this cohort. I-A Related Work Many other works [ 4 , 8 , 5 , 2 , 1 , 6 ] have found that services advertising end-to-end encrypted file storage did not achieve their advertised security properties. These related to a variety of issues including metadata leakage and the opportunity for the server to alter files undetectably, as well as breaches of confidentiality and integrity of stored files. Most of them required more subtle attacks than we describe in this paper. II Store File Figure 2 illustrates the options that the user must complete when storing a file. Figure 3 shows that the password length has now increased from 6 to 8 characters. Fig. 2: Store File Window, before our notification to Chain-FS. Minimum password length is 6. Fig. 3: Store File Process, more than 90 days after our notification to Chain-FS. The minimum password length has been increased to 8. II-A How file storage works To store a file, the user must provide a password in the ‘Encryption key’ field, as shown in Section II . The user-provided key must be between 6 and 32 characters long and must include at least one special character (i.e., a character from the set {!@#$%&̂*}). (Since our first report to Chain-FS, the minimum password length has increased to 8 characters, but this makes no substantial difference to the analysis presented here. See Figure 3 ). As a result, users are permitted to use very short keys such as !!!!!! . The user-provided key is directly used as the AES key without utilizing standard key derivation methods such as PBKDF2. If the user-provided key is shorter than 32 characters, it is padded with zeros to reach the required length. If the key exceeds 32 characters, an error is generated. The selected file is encrypted using AES-256 in CBC mode with PKCS7 padding and an all-zero Initialization Vector (IV), then hashed using SHA-256. Both the encrypted file and its hash value are uploaded to the server. Upon successful storage, the server responds with a unique file GUID. To retrieve the file, the user provides the same password, which is used locally to decrypt the encrypted file downloaded from the server. Figure 4 and Figure 5 summarize the file store and file download processes. Fig. 4: Store File Process. Fig. 5: Download File Process. II-B Required cryptographic corrections It is true that file storage and retrieval is an end-to-end encrypted process, with the file encrypted and decrypted on the client side and only the encrypted version sent to the server. However, there are several vulnerabilities in this design that allow a malicious server to extract various levels of information about encrypted files, including complete decryption in some cases. This section lists the main problems that need to be corrected in the file storage protocol. Unpredictable Initialization Vector (IV). Since the IV is an all-zeros vector, encrypting two identical files with the same password will produce identical ciphertexts. Moreover, if two files share the same first n blocks (that is, 16n bytes), the ciphertexts for the corresponding n blocks will also be identical, if the same password is used. This compromises the security of the encryption and reveals patterns in the data. To resolve this issue, an unpredictable IV must be used. Key length and complexity. The current minimum key length of 6 characters makes brute-force attacks feasible (and the update to 8 does not substantial impact that). The extra restrictions (one special character, etc) actually reduce the search space. A randomly generated password of about 12–14 characters is the minimum required to make exhaustive search attacks impractical. However, it is important to note that brute-force attacks are the least efficient way of guessing passwords. Even with a 14-character minimum, users often create predictable passwords (e.g., their pet’s name followed by a few special characters). Strong emphasis must be placed on the use of properly randomized passwords. Users should be educated about the risks of using simple or reused passwords, as such practices can still allow attackers to decrypt files despite longer key lengths. To address this challenge, integrating password-related fields with trusted password

Read the complete text