Examining Interplay of Compression and Encryption and Applicability to 5G Teleoperations

Large scale networking and device connectivity support the modern world. From phones to pacemakers, devices must communicate critical information over networks quickly, safely, and privately. With the increasing integration of digital components into the economy comes a rise in the amount of cybercrime committed. Malicious actors aim to take control of devices, eavesdrop on data, and in some cases, cause physical harm. With many new devices coming online(Alam, 2018), this concern is quite substantial. Further, these devices often use mobile networks, such as is the case with many IoT devices and teleoperation suites. These networks often have limited resources that must be allocated between the connected devices, leading to each device receiving restrained service quality.

Protecting the data of these devices while also staying within the restrictions of constrained networks involves the use of compression and encryption algorithms. Generally speaking, encryption algorithms secure data with the possibility of increasing its size, and compression algorithms reduce the size of data. These algorithms are often paired together to achieve some security and compression capacity. These algorithms take time to operate, and thus may complicate real-time use cases such as teleoperations. Further, the timing and size impacts of these algorithms change based on the order in which they are applied to a given piece of digital information. The current understanding is that unless there are specific attacks to guard against such as CRIME(att, [n. d.]), one should compress first (CF) rather than encrypt first (EF) to achieve the best compression without compromising security(IBM, 2023; Kumar and Gandhi, 2012; Singh, [n. d.]; Rossevelt, 2023; Fleischhacker et al., 2022). To further resolve this, newer algorithms aim to combine compression and encryption to achieve more harmonious results(Carpentieri, 2018; Kumar and Gandhi, 2012; Chen et al., 2021). While there is substantial support for CF, there is not a robust exploration of the specific performance impacts of the relative order.

In this section we will cover some background concepts and related work for understanding the interactions for compression and encryption. The essential metrics this paper will cover are data size (expressed as file sizes), compression ratio (relative size of data after operations compared to original), and operation time (time taken to compress, encrypt, decrypt, and decompress). However, a core security metric, entropy, is important for measuring the ”disorder or randomness in a closed system”(NIST, [n. d.]). In other words, entropy represents how encrypted data and/or how compressible it is(Veytsman, 2016). Broadly, compression uses patterns for its purposes, and encryption attempts to break patterns(Veytsman, 2016). We note entropy as an important aspect of compression and encryption, but we save an exploration of entropy and operation order for future work.

Broadly, there are two orders to apply compression and encryption: Compression-First (CF) and Encryption-First (EF). CF is the most common approach and is understood to be high performing while also retaining high entropy(Carpentieri, 2018; enc, [n. d.]; Singh, [n. d.]; Rossevelt, 2023). EF is not common, but may be employed to avoid certain attacks that take advantage of the size of encrypted messages such as CRIME(att, [n. d.]; Veytsman, 2016; enc, [n. d.]). Generally speaking, encrypted data that is easily compressed is not well encrypted(IBM, 2023), and alongside the other issues it may present, is also vulnerable to differential cryptanalysis(Chen et al., 2021). Due to the pressure of requiring security while maintaining reasonable data sizes, newer algorithms have emerged to integrate the two operations into a cohesive algorithm(Carpentieri, 2018; Kumar and Gandhi, 2012; Chen et al., 2021). Our work focuses on a set of common individual algorithms and saves the cutting edge algorithms for future work.

To formally evaluate the timing and file size impacts of a given file’s compression and encryption life cycle, we outline a process timeline for operations (encryption or compression) as illustrated in Figure 1. From this process, we highlight three high level metrics: (1) Operation and Operation inverse 1&2 times, which measures the time for a discrete compress/decompress and encrypt/decrypt. (2) Total time, which is the sum total of all operations and inverses for a particular algorithm grouping. Finally, (3) intermediate and final file sizes which provide us insights into the operation-specific impacts and overall impact on the targeted data file.

We examine the effect that changing the order of compression and encryption has on the file size and operation time. Using the process outlined in Fig. 1, we measure the time elapsed at each step of a file being encrypted and compressed, written to disk, and subsequently decrypted and decompressed. We vary the order of compression and encryption. We select 4 common compression algorithms and 3 encryption algorithms (outlined in Table 1 along with 12 different file sizes (Table 2). The files are taken from a selection of vehicle sensors such as LiDAR and GPS unit samples, public large and small text samples such as enwik8(enw, [n. d.]), and a free testing PDF. Tests on files larger than or equal to 10MB are repeated 32 times whereas tests on files smaller than 10MB are repeated 100 times. These repetitions, totaling 39840 tests, alleviate natural variation in operation times and provide a clean average for each combination.

In this section we review the results of our battery of compression and encryption operations against our test files. We first consider the timing impacts of our operations and then consider the impact order has on file size.

From our assessments of algorithm time and size performance, we now apply these observations to a practical use case: teleoperations over 5G.

For real-time transmission, the sum of delay from file read to file reception must be less than 100ms (0.1 seconds)(Nielsen, 1993). For an example with a 10MB PDF file, we add a network latency variable between Operation 2 and Operation 2 Inverse (visualized in Figure 4) to approximate network transmission time. When considering the fastest algorithm for file sizes $<$1MB, we observed files are encrypted, compressed, transmitted with a theoretical and static network latency of 50ms, decrypted, and decompressed in less than 100ms 100% of time. In files with size 1MB, real-time transmission is demonstrated in 73% of samples with the fastest algorithm pair. Lastly, in files $\geq$10MB, the fastest algorithm pair achieves real-time transmission 0% of the time.

Broadly, this means that for larger size data transmissions, the algorithms may not be quite performant enough to apply encryption and compression together, requiring increases in capacity or intelligent decision systems.

This work can be considered a light but robust exploration of the interplay of compression and encryption. Further work can expand on the existing dimensions, including more algorithms and data types. Additionally, examining in detail the impact interplay has on the entropy and other security considerations may be examined as well.

In this paper, we conducted an assessment of the interplay of compression and encryption and its implications on 5G teleoperations. From these examinations, we confirmed that CF is generally the better performing operation order for compression and encryption, but that in certain algorithmic pairs the inverse may be true. Additionally, the impact of using both encryption and compression for files that are sufficiently large will make them unsuitable for real-time transmission, dampening the possibilities for 5G-teleoperations using secure and efficient transmission.