Low-cost Authentication Paradigm for Consumer Electronics Within ...

Low-cost Authentication Paradigm for Consumer Electronics Within the Internet of Wearable Fitness Tracking Applications Fatemeh Tehranipoor Dept. of Electrical and Computer Engineering San Francisco State University San Francisco, CA 94132, USA

Nima Karimian, Paul A. Wortman, and John A. Chandy Dept. of Electrical and Computer Engineering University of Connecticut Storrs, CT 06269, USA

Abstract— Nowadays, the consumer electronic (CE) market has grown to include smart devices for the Internet of Things. With increased adoption of these smart IoT devices, there is a high demand for safe and secure authentication for the purpose of maintaining effective access control. This paper addresses the concerns of authenticating consumer electronic devices using low-cost methods for implementation within customer healthcare frameworks. Our aim is not only to tackle the security issues related to the malicious behavior of supply chain counterfeit devices but also for establishing a secure architecture platform base for users to authenticate within a consumer electronic network. The proposed low-cost solution implements Physical Unclonable Function (PUF) for device verification and incorporates human behavioral characteristics (Biometrics) for user authorization as a comprehensive codesign method.

I. I NTRODUCTION The Internet has drastically changed the way we live, moving interactions between people at a virtual level in several contexts spanning from professional life to social relationships. The origin of the Internet of Things (IoT) can be traced back to the development of the Internet (interconnected network of computer networks). IoT is a novel paradigm that is rapidly gaining ground in the scenario of modern wireless telecommunications [2]. IoT is the interconnected network of objects (things). It is not limited to just computer networks; consumer electronics are becoming incorporated through global access of the Internet. Another basic property of these things (IoT) is push button connectivity to the Internet or peer devices. The IoT continues to grow as consumer electronics become more interconnected. Cisco predicts 50 billions of devices will be connected to the Internet by 2020 [1]. One important growing application of consumer electronic devices is in wearable health tracking solution (healthcare domain). There are benefits provided by IoT technologies to the healthcare domain and the resulting applications can be grouped mostly into: tracking of objects and people (staff and patients), identification and authentication of people, automatic data collection and sensing [3]. The reason we see a movement towards this form of tracking using IoT devices is because now the technology has become more convenient and cost effective for consumers. The author is with the Electrical and Computer Engineering Department, San Francisco State University. ([email protected])

A. Security and Challenges of Internet Connected Wearable Consumer Electronics In order to secure electronic devices, one has to consider security from initial design to the end of life of a given device. Figure 1 illustrates the steps for any consumer electronic device that goes from initial design to deployment. As shown in Figure 1, these attacks can occur in various places along the supply chain. Common issues with early life of electronic devices includes Overproduction (Counterfeiting), Hardware Trojans and Reverse Engineering by malicious parties. Furthermore, current consumer electronics are becoming more and more integrated into the Internet of Things. Due to this adoption, consumer electronic devices (e.g. wearable fitness trackers) must operate in a more resource constrained environments. Authentication within IoT becomes more complicated due to the need for a low resource consuming solution. IoT is extremely vulnerable to attacks for several reasons. First, physical attacks to the unattended components are easy. Second, since most of the communications are wireless, eavesdropping is extremely simple. Third, IoT components cannot implement complex schemes to support security since they are characterized by limited resources and capabilities. B. Problem Statement Consumer electronics have always had the concern of being counterfeited between the fabrication process and vendors storage. To overcome this issue, physical unclonable functions can be used to verify that hardware has not been tampered with. By using this technique, a vendor can be assured that their consumer electronic devices will operate as expected. The next concern for the lifetime of a device is authenticating a user in order to recognize the right individual. Consumer electronics tend to run in a resource constrained environment, therefore any authentication technique needs to be low cost to optimize the device’s lifetime. In order to solve this problem, human behavioral characteristics of individuals (biometrics) is a promising solution for unique customer identification [19]. This paper attempts to balance the concern of both architectural and functional aspects of authenticating consumer electronic devices and their users within a larger network. To better focus the discussion of our work, we examine

Fig. 1: Security Threat Sources in Supply chain Flow for Consumer Electronics

wearable consumer electronic devices (fitness trackers) being used to exchange information and data through in a potentially unsecure communication medium. For the sake of our argument, we further refine the application of our work to the infrastructure of IoT devices in their application to authentication within the healthcare domain. In other words, we investigate the issues of trust and trustworthiness within an environment of unknown devices used by unknown individuals over potentially unsecured networks. The remainder of the paper is organized as follows. The infrastructure security in consumer electronic health tracking are described in Section II. Section III demonstrates our proposed security approach for authentication in consumer electronics healthcare monitoring for device verification and user authentication purposes followed by the data analysis and results. We will discuss in Section IV and finally conclusions are given in Section V. II. I NFRASTRUCTURE S ECURITY IN C ONSUMER E LECTRONIC H EALTH T RACKING D EVICES The many uses of the systems and products that connect to the Internet of Things (IoT) are changing business in numerous industries. Vendors and consumers both stand to benefit from an increased presence of IoT in everyday life (e.g. healthcare, home security). As a result, individuals have adopted these types of consumer electronics for monitoring and collecting/storing data in order to better understand their daily habits. Furthermore, these customers expect their data to be kept safe and secure through trustworthy authentication. One growing application of consumer electronics is fitness tracking which can provide constant and distributed health monitoring within a healthcare IoT network. Some uses of healthcare IoT are mobile medical applications or wearable devices that allow patients to capture their health data. Hospitals use IoT to keep tabs on the location of medical devices, personnel and patients. Consumer electronics can upload the collected vital signals from the sensors (fitness tracking) to the cloud from smartphones, tablet computers, or home remote controllers for a doctor or trained specialist, this data can be processed by signal processing analysis, through the cloud systems in real time across large sets of any patient supplied data. This scenario will reduce the waiting time at emergency rooms and minimize hospital visits. In addition, by analyzing the

patients data in the cloud systems the doctors can respond more quickly and help the patients to increase their quality of life. Moreover, by analyzing the patient’s vital signals from consumer electronic devices, one can use this analysis to help identify and predict unhealthy biological signals like heart attacks for cardiovascular patients; diagnose epilepsy from electroencephalography (EEG) scans and blood pressure detection from photoplethysmogram (PPG). Wearable devices are now at the heart of just about every discussion related to the IoT; a hallmark of the Internet of Things and the most ubiquitous of its implementations to date. Health and fitness oriented wearable devices that offer biometric measurements such as heart rate, perspiration levels, muscle activity, and even complex measurements like oxygen levels in the bloodstream are also becoming available. As wearable IoT use grows and the number of human characteristic signals (biosignals) being recorded and processed increase, one needs to account for the security and privacy concerns of the individuals using the wearable technology. Therefore, risks include possible harm to the patient’s safety and health, loss of protected health information and unauthorized access to devices and exploits of vulnerabilities in the consumer electronic devices. As a result, an authentication framework for internet connected consumer electronics is a necessary requirement in any information system to ensure the availability of information to authorized users only. The users can be authenticated using biometric modalities but the authenticity of the transferring device (hardware) can be questioned. Thus, we present a low-cost authentication framework that binds authentication techniques together by using Physical Unclonable Function (PUF) and ECG biometrics to address both the authentication of devices and users (patients). In fact, our paradigm is able to balance between performance, cost, and reliability. III. P ROPOSED S ECURITY A PPROACH FOR AUTHENTICATION IN C ONSUMER E LECTRONICS H EALTHCARE M ONITORING Figure 2 is a visual representation of the potential connectivity of IoT-based ubiquitous wearable healthcare devices. One will notice that at the center of this mapping is a common cloud where these devices exchange biometric signal information for the purpose of storage, authentication, or for medical observation. As one can see, there are varying

Patient and Bio Signals (Data Provider)

IoT Resource Providers

EEG

Smart Pill Bottles Iris ECG Fingerprint

PUF Biometrics Wireless TPM

Cloud

Zigbee Wired Ethernet Serial Wireless

Smart Phones Computer

EMG Gait Connection Center (Protocols)

Tablet PDA

Laptop

Healthcare Center

Fig. 2: A platform of IoT-based ubiquitous healthcare solutions using secure hardware elements for Authentication.

solutions ranging from the type of protocols used to hardware solutions that provide effective and efficient security and privacy. The authentication mechanisms typically are using passwords, smart cards, tokens, and biometric modalities that are applied to reduce incidents harmful to patients including wrong drug, dose, time and procedure. The biometric features of every human are discriminatively identifiable, nontransferable, and non-reproducible, so why wouldn’t a medical facility use these signals for identification? Especially since they are already collecting this same information from users (patients). Using wearable consumer electronics (sensor nodes) for human monitoring has its own vulnerabilities since the devices may leak personal information [7]. One must address the limitations and security concerns of the underlying hardware to be sure that both patients and staff (different users) are equally protected from malicious, or erroneous, behavior. A. Consumer Electronic Device Verification Counterfeit hardware is a problem that plagues not only larger data mainframes, but also is equally dangerous in the embedded systems market [10]. Generally, the concerns center around security and how to detect hardware Trojans or other malicious ‘motherboard’ behavior. Other concerns include the use of non-standard materials that can cause the fraudulent devices to produce incorrect signals or readings during its operation. As indicated in Figure 1, there is always the threat of reverse engineering or other malicious behavior for the purpose of obtaining sensitive information, cloning device behavior or inserting Trojans. One method of working around this would be to use ‘trusted’ devices already obtained and alter their function to attempt to include security. We can use some of the existing components in the wearable devices to make a security layer on top of the system gathering signals from a patient. But as any security professional will tell you, one cannot plan for security as an afterthought. Therefore, there is a need to authenticate that a given piece of hardware is ‘trustworthy’. Physical Unclonable Functions (PUFs) are one

Fig. 3: Enterprise Biometrics Devices and Licenses by Modality, World Markets [8].

of the hardware solutions commonly used to authenticate with any given device. PUF is a physical structure of a device that is hard to clone due to its inherent, device unique and deep-submicron process variations. PUFs generate unique IDs by exploiting the uncontrollable process variations associated with modern IC fabrication. An ideal PUF takes an input (challenge) and gives a random output (response) which is unique for every device; supplying an exponential number of challengeresponse-pairs. A trusted party, when in possession of an authentic IC, applies randomly chosen challenges to obtain unpredictable responses; which are stored for future authentication operations. To check the authenticity of an IC later, the trusted party selects a challenge that has been previously recorded but has never been used for an authentication check operation. If the PUF response matches (i.e., is close enough to) the previously recorded one, the IC is authentic because only the trusted party should know that challengeresponse-pair. To protect against man-in-the-middle attacks, challenges are never reused. Therefore, the challenges and responses can be sent in the clear during authentication operations [13]. Although there are various types of PUFs, we focus more on memory based PUFs due to the availability of the memories in almost every embedded device [11], [15], [12]. Memories inside the wearable devices can be used for authentication purposes. The cost of including PUFs into existing hardware is relatively small. This technology allows for low-cost authentication of embedded system elements while generating volatile keys for cryptographic applications. As indicated in [11], memory-based PUF such as DRAM PUF has enough reliability, uniqueness and randomness to be used for authentication purposes. We will briefly describe our method for selecting the reliable keys from DRAM PUFs. For the DRAM PUF experiments, we have considered 3 DRAMs (DRAM1, DRAM2, and DRAM3). We were not only evaluated the memories under normal conditions, but also under various environmental conditions as well (voltage variation, temperature variation, and aging). The reason of doing that is because some of the bits might flip (from ’1’ to ’0’ or vice versa) under different conditions. Then, we have tried to find the most

Fig. 4: Security Approach for Low-Cost Verification of Devices and Patients in the healthcare domain. TABLE I: Percentage of bit flips across multiple measurements under normal operating conditions.

n=1 n=2 n=3

DRAM1 BLine Algo 47.3% 2.1% 45.9% 1.5% 45.2% 0.9%

DRAM2 BLine Algo 50.4% 2.3% 46.9% 1.4% 45.7% 0.8%

DRAM3 BLine Algo 49.5% 3.1% 44.2% 1.7% 43.1% 1.1%

stable bits among all possible bits that are available in the memory for creating keys. Note stable bits are the ones that remain unchanged under various conditions. We also have considered the ”neighborhood bit selection” technique which allows to select the key bits among those that have more stable bits around. In general, by considering different approaches, we could select enough reliable bits from DRAM PUFs for authentication. The result of our work is shown in Table I. Table I indicates the percentage of bit flips when reconstructing under just normal conditions. As can be seen, the bit flip rate is reduced to less than 3% with one set of measurements and to less than 1% with 3 sets of measurements. Note that n shows the distinct measurements, Algo is our bit-selection algorithm, and BLine is a random selection in order to compare the result of the Algo with. In our infrastructure, any medical IoT devices used must first be registered with the hospital (acting as the trusted party mentioned earlier). When a patient comes to purchase a medical IoT device from the hospital’s stores, they will perform a device enrollment that will allow the hospital to associate the given device with the specific patient. Once finished with enrollment, the patient will go home where they will activate their medical IoT system. This system will send a PUF response to the hospital database, that the hospital will verify to their enrollment database. Once validated, the hospital will send an enabling signal back to the patient’s medical IoT device allowing it to read their biological signal, thus authenticating the device within the medical IoT network. The patient authentication aspect will be described in the next section. With an expected use of

millions of healthcare IoT devices there is a clear need for being able to authenticate these devices as trustworthy for healthcare related applications. Since these devices will already have some memory along with individualistic deepsubmicron process variations, we propose that PUFs present a simple, low-cost solution for providing a base trust for staff and patients using these IoT devices. Once one trusts the operation of a given IoT device, it can be used for accurate reading of biomedical modalities which can be used to authenticate individuals from a pool of unknown users. Thus, this same technique can be used to authenticate consumer electronic devices. B. Patient Authentication Healthcare applications are a very necessary part of our lives, providing the best medicals facilities to people suffering from various diseases. Hence, it is necessary for the hospitals to keep track of its day to day activities and the records of the incoming patients. Monitoring patients remotely, collecting data and protecting them from tampering is very crucial to address security and privacy concerns. Although there are many benefits of healthcare application, there has always been a threat of hackers getting access to the sensitive information of the patients that can make misuse of that information leading to problems. To reduce preventative caution and impact level of the risk factor, we have investigated biometric authentication. Biometric authentication/identification typically refers to utilizing interior and exterior physiological characteristics of a person (e.g. EEG, ECG, PPG, PCG, iris, fingerprint). The motivation for the extension of biosignals is in their potential uniqueness, universality, and their resistance to spoofing. This can be seen in their increased used as shown in Figure 3. Electrocardiogram (ECG) signal not only is used for the purpose of diagnosis and treatment, but also can be used for biometric authentication and key generation [14]. ECG signals are one of the popular methods for use in authentication systems. This solution has its own advantages in terms of cost, ease of measurement and power usage. Also, it is possible to use ECG signals to generate a key that can be

used for authentication purposes. The key generated from the ECG features was 727 bits in length, has an average entropy of 0.9810, and passed all the 15 NIST statistical tests. Note that this algorithm requires filtering, processing, and quantization of features as overhead for generating an authentication key. Further details of the process can be found in the previously cited paper. The authentication process requires an enrollment of an individual’s ECG features into a database that can later be accessed to positively identify a user. Due to the use of already collected ECG signals, this is a low-cost solution for dealing with authentication of unknown users within a larger healthcare network. Since fitness tracking devices already collect bio-signals, this technique is easy to adapt for consumer electronic authentication. 1) Reliability of Keys: Since Biometric data are noisy in nature, they introduce binary errors during key generation process. Therefore, we have considered ECG-based biometric key generation using quantization technique to eliminate noise, thus achieving maximum reliability and entropy for the produced keys. We have used freely available databases for two biometric modalities, the PTB diagnostic database [18]. Our approach is briefly discussed below which includes two major steps: 1. Preprocessing: In this step, we have considered a Infinite Impulse Response (IIR) filter to eliminate different source noises from heart signal such as Motion Artifact (MA), Baseline Drift (BD), Electromyography (EMG). After filtering process, we have employed feature extraction techniques (e.g. Wavelet transform) in order to generate a pool of various possible features. 2. Bit-extraction: During the bit-extraction/selection, our aim is to transfer the real decimal values into different bit strings of various lengths using the quantization technique. In order to do that, first we need to create normalized Probability Density Function (PDF) from the population. Then, statistical methods (e.g. boundaries, thresholds, helper data) are applied onto each feature to determine boundaries to quantize the feature into one or more key bits. Note that the features that are defined inside a certain boundaries and thresholds will be then selected as a reliable feature, otherwise will be automatically discarded. Furthermore, for achieving the maximum entropy, the boundaries across the PDF are equally distributed. As can be seen in Table II, we achieved 727 as the average key length for normal ECG signals. The minimal values for reliability of keys is about 95% which can become much higher using error correction code (ECC). Note that the average entropy is around 0.99 which is very close to the ideal case value, 1. Our results show that ECG reading has the potential for cryptographic communication and authentication purposes. 2) Randomness of Keys: The NIST statistical test suite is used to evaluate the ”randomness” of the bit strings. Based on this test, we can determine whether a data set has a recognizable pattern or the process that has been generated is significantly random. This test is a statistical package consisting of several types of tests to evaluate the randomness

TABLE II: Reliability and Min-entropy Result based on Biometric Key Generation Biometric modality Average Reliability Minimal Maximal Average Entropy Minimal Maximal Average Key bits

Normal ECG 98.2 94.7 99.9 0.9810 0.864 1.00 727

Fig. 5: NIST Statistical Test Suite Results for Randomness of Keys

of binary sequences. Each statistical test is employed to calculate a P-value that shows the randomness of the given sequences based on that test. If a P-value for a test is determined to be equal to 1, then the sequence appears to have perfect randomness. A p-value >= 0.01 (normally 1%) would means that sequence would be random with a confidence of 99% [17]. The results of applying NIST Tests to the generated keys are shown in Figure 5. Note that the 15 tests are as follows: Frequency Test, Block Frequency, Runs, Longest Run, Cumulative Sums, Rank, FFT, Linear Complexity, Overlapping Template, Non-overlapping Template, Approximate Entropy, Universal Statistical Test, Random Excursions Variant, Random Excursions, and Serial. 3) The Time Estimates for Brute-force Attack: Here, we calculate the time of brute force attack in theory. Brute-force attack is one of the most efficient attack against an algorithm. Based on [16], the time required to crack 727 key bits by brute force attack is 81.64e201 days; approximately 5.38e81 millennia. C. Examining Security of our Paradigm An overview of our security approach is illustrated by Figure 4. This approach is split up between a device verification step and a patient authentication step. We show that initially there is a device verification step where the healthcare IoT device’s trust is established. Making use of existing memory on the IoT device, a DRAM-based PUF challenge-response pair can be used to constitute the trustworthiness of IoT

operation. Should this verification fail then the device cannot be trusted and thus will not be used for reading biological signals. This eliminates the concern that this IoT device would not operate as required in healthcare service which is responsible to verify the IC. The authenticated IC can then be used to read the biological signals and perform the necessary processing to produce a ‘bio-key’. This ‘biokey’ is then enrolled to a local hospital allowing for later authentication of a given patient to the medical staff. This step, of course, requires encryption techniques. This same framework is equally effective at authenticating users in a consumer electronic device network. IV. D ISCUSSION The security and protection of sensitive information and data (e.g. biological signals) within the exemplified healthcare market is a serious and necessary concern. As with any consumer electronics network, one must be able to authenticate the trust of various IoT devices, verify the identity of the users (e.g. patients or staff) within the data network, and ensure that information being exchanged is done in a safe and secure manner. Depending on decisions made by development teams, different protocols or security policies may be implemented to achieve this goal. In the case of our example, we extend these concerns to the growth of medical consumer electronic devices that exist within a patient’s (e.g. user’s) home. We proposed the need for a centralized secure authentication module to provide the base level of trust for any embedded systems used to read and exchange biological signals from the user. Furthermore, these IoT devices could also be authenticated within a larger health monitoring network using the central aggregate computer as a proxy [9]. PUF helps to alleviate issues of communication during authentication since the challenge response pairs would not be used more than once. Resource concerns may require that the central aggregate computer have greater computational capability and storage than the rest of the wearable consumer electronic devices. To overcome the issue of encryption, one could potentially use the bio-keys for encryption of data that would be transferred between an authenticated user and any medical practitioner (e.g. authenticated authority). V. C ONCLUSION In this paper, we presented a low-cost authentication paradigm that can be used to establish trust and authentication of wearable consumer electronic devices within a health monitoring network. Through our solution one would know that the information produced from an IoT is trustworthy, that the individuals accessing information can be properly identified, and that the exchange of sensitive biological signals across a network is secure. We are able to overcome the threat of unknown supply chain alterations and attacks through the implementation of PUFs for hardware verification. We then showed that human characteristic signals (biosignals) can be used to uniquely identify users from a pool of unknown individuals. Together these two solutions allow for the establishment of trustworthy hardware

platforms for the purpose of securely authenticating users for access to sensitive data. As the market for wearable consumer electronic devices continues to expand, vendors will need to adopt low-cost authentication techniques. Concerns of supply chain safety require validation of hardware to prevent loss of consumer credentials and data. Our solution helps to solve this issue in a lightweight and easily adoptable framework that can work with any form of consumer electronics. R EFERENCES [1] Cisco. Internet of things (IoT) [Internet]. Available from: http://www.cisco.com/web/solutions/trends/iot/portfolio.html [2] L. Atzori et al., “The internet of things: A survey,” Computer networks, vol. 54, no. 15, pp. 2787–2805, 2010. [3] A. Vilamovskaet al., “Rfid application in healthcare–scoping and identifying areas for rfid deployment in healthcare delivery,” RAND Europe, February, 2009. [4] H. Luo et al., “A remote markerless human gait tracking for ehealthcare based on content-aware wireless multimedia communications,” IEEE Wireless Communications, vol. 17, no. 1, pp. 44–50, 2010. [5] A. Kulkarni et al., “Healthcare applications of the internet of things: A review,” International Journal of Computer Science and Information Technologies, vol. 5, no. 5, pp. 6229–32, 2014. [6] F. Fernandez et al., “Opportunities and challenges of the internet of things for healthcare: Systems engineering perspective,” in Wireless Mobile Communication and Healthcare (Mobihealth), 2014 EAI 4th International Conference on. IEEE, 2014, pp. 263–266. [7] S. Amyx, “Privacy dangers of wearables and the internet of things,” Managing Security Issues and the Hidden Dangers of Wearable Technologies, p. 131, 2016. [8] Tractica, “IT Security Applications Will Drive Growth in Enterprise Adoption of Biometrics Technology During the Next 10 Years,” https://www.tractica.com/newsroom/press-releases/it-securityapplications-will-drive-growth-in-enterprise-adoption-of-biometricstechnology-during-the-next-10-years/, 2015. [9] P Wortman et al., ”Proposing a modeling framework for minimizing security vulnerabilities in IoT systems in the healthcare domain.” In Biomedical & Health Informatics (BHI), 2017 IEEE EMBS International Conference on, pp. 185-188. IEEE, 2017. [10] M. Cavalleri et al.,“A wearable device for a fully automated in-hospital staff and patient identification.” in Conference proceedings:... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference, vol. 5, 2003, pp. 3278–3281. [11] F Tehranipoor et al., ”DRAM-Based Intrinsic Physically Unclonable Functions for System-Level Security and Authentication.” IEEE Transactions on Very Large Scale Integration (VLSI) Systems 25.3 (2017): 1085-1097. [12] C. Herder et al., “Physical unclonable functions and applications: A tutorial,” Proceedings of the IEEE, vol. 102, no. 8, pp. 1126–1141, Aug 2014. [13] G. E. Suh et al., “Physical unclonable functions for device authentication and secret key generation,” in Proceedings of the 44th annual Design Automation Conference. ACM, 2007, pp. 9–14. [14] N. Karimian et al., “Highly reliable key generation from electrocardiogram (ecg),” IEEE Transactions on Biomedical Engineering, vol. PP, no. 99, pp. 1–1, 2016. [15] F. Tehranipoor et al., ”DRAM based intrinsic physical unclonable functions for system level security.” In Proceedings of the 25th edition on Great Lakes Symposium on VLSI, pp. 15-20. ACM, 2015. [16] S. Bruce, “Applied cryptography,” 2nd John Wiley and Sons, Inc, 1996. [17] Rukhin et al., A statistical test suite for random and pseudorandom number generators for cryptographic applications, DTIC Document, Tech. Rep., 2001. [18] A. L. e. a. Goldberger. Physiobank, physiotoolkit, and physionet components of a new research resource for complex physiologic signals. 101(23):e215e220, 2000. [19] N. Karimian et al., ”Evolving authentication design considerations for the internet of biometric things (IoBT).” In Hardware/Software Codesign and System Synthesis (CODES+ ISSS), 2016 International Conference on, pp. 1-10. IEEE, 2016.