Source-linked AI summary
On the Feasibility of Attribute-Based Encryption on Internet of Things Devices
Moreno Ambrosin, Arman Anzanpour, Mauro Conti, Tooska Dargahi, Sanaz Rahimi Moosavi, Amir M. Rahmani, Pasi Liljeberg
TL;DR
The literature lacks a proper assessment of ABE efficiency on resource-constrained IoT devices, despite its use in distributed environments. The authors implement original KP-ABE and CP-ABE and evaluate encryption and decryption across four IoT boards, security levels, and attribute counts. At 80-bit security and 30 attributes, CP-ABE encryption takes ≈5 sec on Raspberry Pi 1; decryption takes ≈2.9 sec on Raspberry Pi Zero; Intel Galileo takes ≈15 and ≈13 sec, respectively.
Problem
The literature lacks a proper assessment of ABE efficiency on resource-constrained IoT devices, despite its use in distributed environments.
Source: §1 · p. 2Method
The authors implement original KP-ABE and CP-ABE and evaluate encryption and decryption across four IoT boards, security levels, and attribute counts.
Source: §1 · p. 2, §3.1 · p. 5Results
At 80-bit security and 30 attributes, CP-ABE encryption takes ≈5 sec on Raspberry Pi 1; decryption takes ≈2.9 sec on Raspberry Pi Zero; Intel Galileo takes ≈15 and ≈13 sec, respectively.
Source: §3.2 · p. 6Takeaways & Limitations
The authors conclude that ABE is feasible on the tested IoT devices, with latency and energy trade-offs informing choices of security level and attribute count.
Source: §3 · p. 3, §3.2 · p. 6, §5 · p. 11Because of space limitations, the paper reports CP-ABE results only, while stating that KP-ABE experiments showed similar quantitative behavior.
Source: §1 · p. 2
Abstract
from arXiv · showhide
Attribute-Based Encryption (ABE) could be an effective cryptographic tool for the secure management of Internet-of-Things (IoT) devices, but its feasibility in the IoT has been under-investigated thus far. This article explores such feasibility for well-known IoT platforms, namely, Intel Galileo Gen 2, Intel Edison, Raspberry Pi 1 Model B, and Raspberry Pi Zero, and concludes that adopting ABE in the IoT is indeed feasible.
1 Introduction
IoT devices create security and privacy challenges when sensitive data crosses untrusted networks or collaborating systems. The paper addresses the limited assessment of ABE efficiency on resource-constrained devices by implementing and evaluating ABE on four IoT boards.
1 Introduction
- IoT deployments can expose sensitive data to untrusted networks, clouds, or collaborating entities, motivating efficient authentication and fine-grained access control.Distributed IoT systems may also need devices to communicate directly across organizational or system boundaries.Source: §1 · p. 1
- The literature lacks a proper assessment of ABE efficiency on resource-constrained IoT devices, despite ABE's use in distributed environments.The paper implements KP-ABE and CP-ABE on four boards and reports CP-ABE results, noting similar quantitative behavior in KP-ABE experiments.Source: §1 · p. 2
2 Expressive Encryption with ABE
ABE applies high-level access policies to encryption and decryption, with KP-ABE and CP-ABE assigning policies to different parts of the exchange. The section explains their policy mechanics and the computational costs associated with attributes.
2 Expressive Encryption with ABE
- KP-ABE associates an access policy with each user's key, while CP-ABE places the policy on the encrypted data and associates attributes with user keys.Decryption is possible when the ciphertext's attributes satisfy the KP-ABE key policy, or when a user's attributes satisfy the CP-ABE ciphertext policy.Source: §2 · pp. 2–3
- CP-ABE encryption requires two exponentiations per policy attribute, while decryption is bounded by l exponentiations and 2l pairings; KP-ABE decryption uses l pairings.Here, l is the number of attributes matching the relevant policy.Source: §2 · p. 3
- Numeric attributes can make policies more expressive, such as restricting access to device models released after a specified date, but may be expensive.The paper examines numeric attributes alongside string attributes.Source: §2 · p. 3
3 Feasibility of ABE on IoT Devices
The study assesses ABE feasibility on resource-constrained IoT boards using latency as its primary criterion. It finds substantial trade-offs among device, security level, and policy size, and examines numeric-attribute policy costs.
3 Feasibility of ABE on IoT Devices
- The study implements original KP-ABE and CP-ABE schemes on representative middle-class IoT devices, retaining the original schemes for compatibility, comparability, and available libraries.It reports CP-ABE results because of space limitations and says KP-ABE experiments showed similar quantitative behavior.Source: §3.1 · p. 5
3.1 Experimental Setup
- Feasibility is judged primarily by latency, which directly impacts consumed energy; cryptographic-operation performance was assessed on four boards at three security levels and policy sizes of 1–30 attributes.The boards were Intel Edison (1335.84 mW baseline power), Intel Galileo Gen 2 (7021.44 mW), Raspberry Pi 1 Model B (2358.4 mW), and Raspberry Pi Zero (1504 mW).Source: §3 · p. 3, §3.1 · p. 5
3.2 Evaluation and Discussion
- At 80-bit security with 30 attributes, encryption takes ≈5 seconds on Raspberry Pi 1 and ≈15 seconds on Intel Galileo; decryption takes ≈3.6 seconds on Pi 1 and ≈2.9 seconds on Pi Zero, versus ≈13 seconds on Galileo.Source: §3.2 · p. 6
- On Intel Edison, encryption spends almost 91% of its time in element_from_hash and element_pow_zn, while decryption spends almost 97% in pairing_apply.These function-level measurements identify where the implementation's operation time is concentrated.Source: §3.2 · p. 7
3.3 Numeric Attributes in ABE
- Numeric policies translate into access trees whose complexity depends on the number's minimum byte representation, with powers of two producing simpler trees and faster encryption.For example, A<256 takes about 1.941 seconds with eleven leaves and two AND gates, while A<32768 takes about 0.547 seconds with three leaves and one AND gate.Source: Figure 4 · p. 9, §3.3 · p. 10
4 Use Case: IoT in Healthcare
In the healthcare prototype, an Intel Edison encrypts sensor files with CP-ABE and sends them over Wi-Fi; at 80-bit security, latency remains below or close to the one-second bound with up to five attributes.
4 Use Case: IoT in Healthcare
- At 80-bit security, overall latency remains below or close to the one-second upper bound with up to 5 attributes.The study treats roughly one second as the latency upper bound for this use case.Source: §4 · p. 11
- An Intel Edison prototype reads medical sensor measurements, encrypts files with CP-ABE, and sends them by Wi-Fi to a server that may be untrusted.Separate on-board services read sensor data and encrypt and transmit the resulting files.Source: §4 · p. 11
- The evaluation focuses on ECG data at approximately 1500 Bytes/sec; UDP transmission takes about 80 msec per packet and AES file encryption averages 45 msec.CP-ABE operations are reported as the most expensive in this setup.Source: §4 · p. 11
5 Conclusion
The authors report ABE feasibility on representative IoT systems and identify software and hardware optimizations for improving efficiency.
5 Conclusion
- The study reports that adopting ABE is feasible in representative IoT systems.The authors suggest the results can serve as a reference for researchers and designers of ABE-based security solutions.Source: §5 · p. 11
- Library efficiency could be improved through memory management, customized data structures, and simplified cryptographic arithmetic tailored to input attributes.Source: §5 · p. 11
- Because CP-ABE and KP-ABE complexity depends on exponentiations and pairings, future work could move operations such as exponentiation to hardware accelerators to improve energy efficiency and execution time.Source: §5 · p. 11
- Figure 5 covers healthcare use-case parameters and latency evaluation on an Intel Edison board at 80-bit security.Source: Figure 5 · p. 12