Understanding the Security Risks of Docker Hub

Abstract

Docker has become increasingly popular because it provides efficient containers that are directly run by the host kernel. Docker Hub is one of the most popular Docker image repositories. Millions of images have been downloaded from Docker Hub billions of times. However, in the past several years, a number of high-profile attacks that exploit this key channel of image distribution have been reported. It is still unclear what security risks the new ecosystem brings. In this paper, we reveal, characterize, and understand the security issues with Docker Hub by performing the first large-scale analysis. First, we uncover multiple security-critical aspects of Docker images with an empirical but comprehensive analysis, covering sensitive parameters in run-commands, the executed programs in Docker images, and vulnerabilities in contained software. Second, we conduct a large-scale and in-depth security analysis against Docker images. We collect 2,227,244 Docker images and the associated meta-information from Docker Hub. This dataset enables us to discover many insightful findings. (1) run-commands with sensitive parameters expose disastrous harm to users and the host, such as the leakage of host files and display, and denial-of-service attacks to the host. (2) We uncover 42 malicious images that can cause attacks such as remote code execution and malicious cryptomining. (3) Vulnerability patching of software in Docker images is significantly delayed or even ignored. We believe that our measurement and analysis serves as an important first-step study on the security issues with Docker Hub, which calls for future efforts on the protection of the new Docker ecosystem.

A Appendix

1.1 A.1 User Study on Sensitive Parameters

We design an online questionnaire that contains questions including “Do you try to fully understand every parameter of the run-commands provided on the Docker Hub website before running those commands?”, “Do you make a security analysis of the compose.yml file before running the image?”, etc. Our questionnaire was sent to our colleagues and classmates, and further spread by them. In order to ensure the authenticity and objectivity of the investigation results, we did not tell any respondents the purpose of this survey. We plan to conduct the user study in the official community of Docker Hub in the future.

Finally, we collected 106 feedback offered by 106 users from various cities in different countries. All of them have benefited from Docker Hub, i.e., they have experiences in using images from Docker Hub. Besides, they are from a broad range from both academia and industry fields, including students and researchers from various universities, software developers and DevOps engineers from different companies, etc.

As described in Sect. 4.2, the results of our user study show that 97% of users only care about if they can successfully run the image while ignoring how the images run, not to mention the sensitivity parameters in run-command and docker-compose.yml file. Even for 68 users who have a background in security research, only 10% of them indicate that they prefer to figure out the meaning of the parameters in run-commands.

1.2 A.2 Novel Attacks Exploiting Sensitive Parameters

Obtaining the Display of the Host. is one of the most powerful parameters provided by Docker, which may pose a serious threat to users. When the operator uses command , the container will gain access to all the devices on the host. Under this scenario, the container can do almost anything with no restriction, which is extremely dangerous to the security of users. More specifically, allows a container to mount a partition on the host. By taking a step further, the attacker can access all the user files stored on this partition. In addition to accessing user files, we design an attack to obtain the display of a user’s desktop. In fact, with , a one-line code, , is sufficient for attackers to access user display data. Furthermore, by leveraging simple image processing software  [7], attackers can see the user’s desktop as if they were sitting in front of the user’s monitor.

Spying the Process Information on the Host. is a parameter related to namespaces. Providing allows a container to share the host’s PID namespace. In this case, if the container is under the control of an attacker, all the programs running on the user’s host will become visible to the attacker inside the container. Then, the attacker can utilize these exposed information such as the PID, the owner, the path of the corresponding executable file and the execution parameters of the programs, to conduct effective attacks.

1.3 A.3 Case Study of Malicious Images

We manually conduct analysis on detected malicious images. For instance, the image on Docker Hub is detected as malicious by our framework. The entry-file of this image is   [7]. According to the name and entry-file of the image, the functionality of this image should be image and video processing. However, our framework detects that the real functionality of the entry-file is mining Bit-coins. By leveraging the syscall log reported by our framework, we determine that the real identity of this image is a Bit-coin miner. Thus, once users run the image, their machines will become slaves for cryptomining.

Fig. 5.

Vulnerabilities existing in the latest images.

1.4 A.4 Distribution of Vulnerabilities

We investigate the distribution of vulnerabilities in the latest version of all official images. First, we observe that the latest official images contain 30,000 CVE vulnerabilities. Figure 5(a) categorizes these CVE vulnerabilities into 6 groups according to the severity levels assessed by the latest CVSSv3 scoring system  [12]. Although only 6% of vulnerabilities are highly/critically severe, they exist in almost 30% of the latest official images. Furthermore, we conduct a similar analysis on the latest images in the 10,000 most popular community repositories. As shown in Fig. 5(b), the ratios of vulnerabilities with medium and high severity increase to over 37% and 8%, respectively, which are higher than those of official images. In addition, it is quite alarming that more than 64% of community images are affected by highly/critically severe vulnerabilities such as the denial of service and memory overflow. These results demonstrate that both official and community images suffer from serious software vulnerabilities. Additionally, community images contain more vulnerabilities with higher severity. Hence, we propose that software vulnerability is an urgent problem which seriously affects the security of Docker images.