Open Access

Modern society is more than ever striving for digital connectivity -- everywhere and at any time, giving rise to megatrends such as the Internet of Things (IoT). Already today, 'things' communicate and interact autonomously with each other and are managed in networks. In the future, people, data, and things will be interlinked, which is also referred to as the Internet of Everything (IoE). Billions of devices will be ubiquitously present in our everyday environment and are being connected over the Internet. As an emerging technology, printed electronics (PE) is a key enabler for the IoE offering novel device types with free form factors, new materials, and a wide range of substrates that can be flexible, transparent, as well as biodegradable. Furthermore, PE enables new degrees of freedom in circuit customizability, cost-efficiency as well as large-area fabrication at the point of use. These unique features of PE complement conventional silicon-based technologies. Additive manufacturing processes enable the realization of many envisioned applications such as smart objects, flexible displays, wearables in health care, green electronics, to name but a few. From the perspective of the IoE, interconnecting billions of heterogeneous devices and systems is one of the major challenges to be solved. Complex high-performance devices interact with highly specialized lightweight electronic devices, such as e.g. smartphones and smart sensors. Data is often measured, stored, and shared continuously with neighboring devices or in the cloud. Thereby, the abundance of data being collected and processed raises privacy and security concerns. Conventional cryptographic operations are typically based on deterministic algorithms requiring high circuit and system complexity, which makes them unsuitable for lightweight devices. Many applications do exist, where strong cryptographic operations are not required, such as e.g. in device identification and authentication. Thereby, the security level mainly depends on the quality of the entropy source and the trustworthiness of the derived keys. Statistical properties such as the uniqueness of the keys are of great importance to precisely distinguish between single entities. In the past decades, hardware-intrinsic security, particularly physically unclonable functions (PUFs), gained a lot of attraction to provide security features for IoT devices. PUFs use their inherent variations to derive device-specific unique identifiers, comparable to fingerprints in biometry. The potentials of this technology include the use of a true source of randomness, on demand key derivation, as well as inherent key storage. Combining these potentials with the unique features of PE technology opens up new opportunities to bring security to lightweight electronic devices and systems. Although PE is still far from being matured and from being as reliable as silicon technology, in this thesis we show that PE-based PUFs are promising candidates to provide key derivation suitable for device identification in the IoE. Thereby, this thesis is primarily concerned with the development, investigation, and assessment of PE-based PUFs to provide security functionalities to resource constrained printed devices and systems. As a first contribution of this thesis, we introduce the scalable PE-based Differential Circuit PUF (DiffC-PUF) design to provide secure keys to be used in security applications for resource constrained printed devices. The DiffC-PUF is designed as a hybrid system architecture incorporating silicon-based and inkjet-printed components. We develop an embedded PUF platform to enable large-scale characterization of silicon and printed PUF cores. In the second contribution of this thesis, we fabricate silicon PUF cores based on discrete components and perform statistical tests under realistic operating conditions. A comprehensive experimental analysis on the PUF security metrics is carried out. The results show that the silicon-based DiffC-PUF exhibits nearly ideal values for the uniqueness and reliability metrics. Furthermore, the identification capabilities of the DiffC-PUF are investigated and it is shown that additional post-processing can further improve the quality of the identification system. In the third contribution of this thesis, we firstly introduce an evaluation workflow to simulate PE-based DiffC-PUFs, also called hybrid PUFs. Hereof, we introduce a Python-based simulation environment to investigate the characteristics and variations of printed PUF cores based on Monte Carlo (MC) simulations. The simulation results show, that the security metrics to be expected from the fabricated devices are close to ideal at the best operating point. Secondly, we employ fabricated printed PUF cores for statistical tests under varying operating conditions including variations in ambient temperature, relative humidity, and supply voltage. The evaluations of the uniqueness, bit aliasing, and uniformity metrics are in good agreement with the simulation results. The experimentally determined mean reliability value is relatively low, which can be explained by the missing passivation and encapsulation of the printed transistors. The investigation of the identification capabilities based on the raw PUF responses shows that the pure hybrid PUF is not suitable for cryptographic applications, but qualifies for device identification tasks. The final contribution is to switch to the perspective of an attacker. To judge on the security capabilities of the hybrid PUF, a comprehensive security analysis in the manner of a cryptanalysis is performed. The analysis of the entropy of the hybrid PUF shows that its vulnerability against model-based attacks mainly depends on the selected challenge building method. Furthermore, an attack methodology is introduced to assess the performances of different mathematical cloning attacks on the basis of eavesdropped challenge-response pairs (CRPs). To clone the hybrid PUF, a sorting algorithm is introduced and compared with commonly used supervised machine learning (ML) classifiers including logistic regression (LR), random forest (RF), as well as multi-layer perceptron (MLP). The results show that the hybrid PUF is vulnerable against model-based attacks. The sorting algorithm benefits from shorter training times compared to the ML algorithms. If the eavesdropped CRPs are erroneous, the ML algorithms outperform the sorting algorithm.

A physical unclonable function (PUF) is a hardware circuit that produces a random sequence based on its manufacturing-induced intrinsic characteristics. In the past decade, silicon-based PUFs have been extensively studied as a security primitive for identification and authentication. The emerging field of printed electronics (PE) enables novel application fields in the scope of the Internet of Things (IoT) and smart sensors. In this paper, we design and evaluate a printed differential circuit PUF (DiffC-PUF). The simulation data are verified by Monte Carlo analysis. Our design is highly scalable while consisting of a low number of printed transistors. Furthermore, we investigate the best operating point by varying the PUF challenge configuration and analyzing the PUF security metrics in order to achieve high robustness. At the best operating point, the results show areliability of 98.37% and a uniqueness of 50.02%, respectively. This analysis also provides useful and comprehensive insights into the design of hybrid or fully printed PUF circuits. In addition, the proposed printed DiffC-PUF core has been fabricated with electrolyte-gated field-effect transistor technology to verify our design in hardware.

Objectives: Speech recognition on the telephone poses a challenge for patients with cochlear implants (CIs) due to a reduced bandwidth of transmission. This trial evaluates a home-based auditory training with telephone-specific filtered speech material to improve sentence recognition. Design: Randomised controlled parallel double-blind. Setting: One tertiary referral centre. Participants: A total of 20 postlingually deafened patients with CIs. Main outcome measures: Primary outcome measure was sentence recognition assessed by a modified version of the Oldenburg Sentence Test filtered to the telephone bandwidth of 0.3-3.4 kHz. Additionally, pure tone thresholds, recognition of monosyllables and subjective hearing benefit were acquired at two separate visits before and after a home-based training period of 10-14 weeks. For training, patients received a CD with speech material, either unmodified for the unfiltered training group or filtered to the telephone bandwidth in the filtered group. Results: Patients in the unfiltered training group achieved an average sentence recognition score of 70.0%±13.6% (mean±SD) before and 73.6%±16.5% after training. Patients in the filtered training group achieved 70.7%±13.8% and 78.9%±7.0%, a statistically significant difference (P=.034, t10 =2.292; two-way RM ANOVA/Bonferroni). An increase in the recognition of monosyllabic words was noted in both groups. The subjective benefit was positive for filtered and negative for unfiltered training. Conclusions: Auditory training with specifically filtered speech material provided an improvement in sentence recognition on the telephone compared to training with unfiltered material.

The ability to detect a target signal masked by noise is improved in normal-hearing listeners when interaural phase differences (IPDs) between the ear signals exist either in the masker or in the signal. To improve binaural hearing in bilaterally implanted cochlear implant (BiCI) users, a coding strategy providing the best possible access to IPDs is highly desirable. Outcomes of a previous study (Zirn, Arndt et al. 2016) revealed that a subset of BiCI users showed improved IPD detection thresholds with the fine structure processing strategy FS4 compared to the constant rate strategy HDCIS using narrowband stimuli. In contrast, little differences between the coding strategies were found for broadband stimuli with regard to binaural speech intelligibility level differences (BILD) as an estimate of binaural unmasking. Compared to normalhearing listeners (7.5 ± 1.2 dB) BILD were small in BiCI users (around 0.5 dB with both coding strategies). In the present work, we investigated the influence of binaural fitting parameters on BILD. In our cohort of BiCI users many were implanted with electrode arrays differing in length left versus right. Because this length difference typically corresponded to the distance of two electrode contacts the first modification of bilateral fitting was a tonotopic adjustment by deactivation of the most apical electrode contact on the side with the deeper inserted array (tonotopic approach). The second modification was the isolation of the residual, most apical electrode contacts by deactivation of the basally adjacent electrode contact on each side (tonotopic sparse approach). Applying these modifications, BILD improved by up to 1.5 dB.

Objective: The present study evaluated whether the poorer baseline performance of cochlear implant (CI) users or the technical and/or physiological properties of CI stimulation are responsible for the absence of masking release. Design: This study measured speech reception thresholds (SRTs) in continuous and modulated noise as a function of signal to noise ratio (SNR). Study sample: A total of 24 subjects participated: 12 normal-hearing (NH) listeners and 12 subjects provided with recent MED-EL CI systems. Results: The mean SRT of CI users in continuous noise was −3.0 ± 1.5 dB SNR (mean ± SEM), while the normal-hearing group reached −5.9 ± 0.8 dB SNR. In modulated noise, the difference across groups increased considerably. For CI users, the mean SRT worsened to −1.4 ± 2.3 dB SNR, while it improved for normal-hearing listeners to −18.9 ± 3.8 dB SNR. Conclusions: The detrimental effect of fluctuating maskers on SRTs in CI users shown by prior studies was confirmed by the current study. Concluding, the absence of masking release is mainly caused by the technical and/or physiological properties of CI stimulation, not just the poorer baseline performance of many CI users compared to normal-hearing subjects. Speech understanding in modulated noise was more robust in CI users who had a relatively large electrical dynamic range.