A Survey on LoRa Networking: Research Problems, Current Solutions and Open Issues

2 A Surv e y on LoRa Netw orking: Research Problems, Current Solutions and Open Issues Jothi Prasanna Shanmuga Sundaram, Student member , IEEE , W an Du, Member , IEEE , Zhiwei Zhao, Member , IEEE Abstract — Wireless networks ha ve been widely deployed for many Internet-of-Things (IoT) applications, like smart cities and precision agricultur e. Low Po wer Wide Ar ea Networking (LPW AN) is an emerging IoT networking paradigm to meet three k ey requir ements of IoT applications, i.e., low cost, large scale deployment and high energy efﬁciency . Among all available LPW AN technologies, LoRa networking has attracted much attention from both academia and industry , since it speciﬁes an open standard and allows us to build autonomous LPW AN networks without any third-party infrastructure. Many LoRa networks have been developed recently , e.g., managing solar plants in Carson City , Nevada, USA and power monitoring in L y on and Grenoble, France. However , ther e are still many resear ch challenges to dev elop practical LoRa networks, e.g., link coordination, resour ce allocation, reliable transmissions and security . This article pr ovides a comprehensive survey on LoRa networks, including the technical challenges of deploying LoRa networks and recent solutions. Based on our detailed analysis of current solutions, some open issues of LoRa networking are discussed. The goal of this survey paper is to inspire more works on improving the performance of LoRa networks and enabling more practical deployments. Index T erms —The Internet-of-Things, Low Power ed Wide Area Networking, LoRa, taxonomy . I . I N T RO D U C T I O N T HE Internet-of-Things (IoT) applications [1], [2], like smart homes and smart cities, become more and more pervasi v e, which result in increasing density and scale of networked sensor deployments [3]–[5]. Ericsson mobility re- port [6] states that connected IoT devices will grow from sev en billion in 2017 to 20 billion in 2023, corresponding to an annual growth rate of 19 % . The IoT applications employ things with sensing capabilities to sense the environment, communicate with other de vices and humans and make intelli- gent decisions. T o connect IoT de vices, wireless networks are required to provide robust operations and wider cov erage with high energy efﬁcienc y [1]. The IoT end devices are mostly battery powered. They are expected to work for a longer span of ﬁ ve to ten years without an y maintenance. These IoT end- devices are also expected to cover a large geographical area. For example, the forest monitoring application installs end devices throughout the forest region. The devices communi- W an Du and Jothi Prasanna Shanmuga Sundaram are with the Department of Computer Science and Engineering, the Uni versity of California, Merced. E-mail: { wdu3, jshanmugasundaram } @ucmerced.edu, Zhiwei Zhao is with the School of Computer Science and Engineering, Uni versity of Electronic Science and T echnology of China. Email: zzw@uestc.edu.cn. W an Du is the ﬁrst corresponding author of this article and Zhiwei Zhao is the second corresponding author . cate small payloads to con vey interesting data like humidity , temperature and other v ariables over a longer distance in a multi-hop manner . The abo ve requirements ha ve led to a ne w branch of IoT net- working technology , called Lo w Po wer W ide Area Networking (LPW AN), as con ventional IoT networking technologies like Zigbee and Bluetooth can only provide a shorter range [2], [7], [8]. LPW AN employs simple netw ork topology and long distance communication with low data rates to attain high energy efﬁcienc y [9]. Existing LPW AN technologies can be divided into three categories i.e., networks based on cellular infrastructure [10], [11], networks using third-party infrastruc- ture [12], autonomous LPW AN networks without any third- party infrastructure [13]. First, e xisting cellular technology covers a wide area but its energy ef ﬁciency does not match LPW AN requirements as the y were not commissioned for machine-type communications. As cellular networks are already densely populated, a new massiv e wave of IoT devices cannot be handled as it leads to heavy interference. T o ov ercome these challenges, intensi ve research is being conducted on Cellular -IoT technologies like L TE-M [14], [15], NB-IoT [10], [11] and EC-GSM [11]. For example, NB-IoT [10], [11] operates at licensed Long-T erm Evolution (L TE) bands using Single-Carrier Frequency Divi- sion Multiple Access (SC-FDMA) for uplink and Orthogonal Frequency Division Multiple Access (OFDMA) for downlink modulation. It facilitates higher Quality-of-Service (QoS) [16]. Second, some service providers, lik e SigFox [12], In- genu [17] and W eightless [18], are proprietary networks. Ingenu [17] is a founding member of IEEE 802.15.4K task group. It leans on completing its stack whereas SigF ox and LoRa focus on faster time to market. It operates at the 2.4GHz band. Ingenu uses Random Phase Multiple Access (RPMA) modulation which giv es higher link budget and cov erage while energy efﬁciency becomes a downside. Ingenu also suffers from interference of other technologies like W iFi, lo w structural penetration of signals and increased propagation loss at high frequencies [19]. SigFox [12] is more popular in European region because of the traction made by widely av ailable vendors like Axom, T exas Instruments and Silicon labs. 100Hz bandwidth (BW) and Ultra-Narrow Band technol- ogy are utilized for transmitting smaller pack ets (12 bytes, up to 140 messages per day) at low data rates (up to 100 bits per second) modulated with Binary Phase Shift Ke ying (BPSK). Major limitations of SigFox includes (i) being proprietary closed source technology , (ii) Low security mechanisms and (iii) restrictions on downlink transmissions [20]. 3 T ABLE I: Comparison of LPW AN technologies LoRa Ingenu Sigfox NB-IoT Third-party infrastructure Open source Closed source Closed source Open Source Operating Band ISM Sub-GHz ISM 2.4GHz ISM Sub-GHz Licensed L TE band 180KHz Channels Multiple SF with 64+8 UL and 8 DL 40 1-MHz channels, 1200 signals per second 360 channels 3 DL and 2 UL Modulation CSS, FSK RPMA-DSSS, CDMA DBPSK, GFSK OFDMA, SC-FDMA Data rate 0.3-37.5 Kbps 78 Kbps UL, 19.5 Kbps DL 100 bps UL, 600 bps DL up to 250kbps Communication Range 5Km -15Km [9], [10] 15Km [9], [10] 1-Km to 5-Km [9], [10] up to 35Km [9], [10] Payload Length up to 250Bytes 10 Kilobytes 12Bytes UL and 8B DL 1600 Bytes Authentication Symmetrical Authentication key Mutual Authentication Burnt-in symmetrical authentication key Mutual Authentication Encryption AES 128bit AES 256bit 5 L TE encryption Finally , LoRa networking [21] is widely used for LPW AN applications because, LoRa networking is an open-sour ce technology that enables autonomous network set-up at low cost. LoRa networks hav e been widely deployed for many applications and research systems. The openness of LoRa makes it an excellent choice for diverse IoT deployments [13]. General IoT applications include smart buildings [22], smart cities [23], smart agriculture [24], smart meters [25], [26] and water quality measurement [27]–[30]. Major LPW AN technologies are compared in T able I. Acronyms used in this table are described in T able II. W orking in sub-GHz band using CSS modulation makes LoRa tech- nology immune to interference as the chirp signal varies its frequency linearly with time. The chirp signals utilize the av ailable bandwidth instantaneously consuming low power than the other LPW AN technologies. A nominal coverage of 5Km-15Km [10] is obtained with higher payload (up to 250Bytes) when compared to other technologies. LoRa networks of fer better downlink capabilities than Sigfox and Ingenu. LoRa networking provides light-weight encryption and authentication mechanisms that can be conﬁgured during activ ation. Another important advantage of LoRa networks is that the conﬁguration and ﬁrmware updates can be sent over the air [31]. Why a new survey on LoRa networking? Raza et. al ascertain the need of LPW AN by justifying the inability of legac y wireless systems to comply with the constraints of LPW AN [9]. The design goals of LPW AN along with various techniques to achieve these goals are discussed. On discussing the challenges and research directions, the authors ﬁnd that most of the working groups focus on PHY and MA C layers. W e argue that the upper layers should also be discussed such as the ef ﬁcient deplo yment of LPW AN. A brief description is pro- vided on technical speciﬁcations of all LPW AN technologies while recent performance measurements, research challenges and solutions are not discussed in detail. Sinha et al. study two leading LPW AN technologies, LoRa and NB-IoT , by comparing their ph ysical features, MA C proto- T ABLE II: Acron yms found in this paper Acronym Description bps bits per second CDMA Code Division Multiple Access CR Code Rate CSS chirps Spread Spectrum DBPSK Differential Binary Phase Shift K eying DL Downlink DSSS Direct Sequence Spread Spectrum EC-GSM Extended Coverage-GSM FSK Frequenc y Shift Ke ying GFSK Gaussian Frequency Shift Ke ying Kbps Kilo bits per second LPW AN Low Power W ide Area Networks NB-IoT Narrow Band-IoT OFDM Orthogonal Frequency Di vision Multiple Access RPMA Random Phase Multiple Access SC-FDMA Single Carrier-Frequenc y Division Multiple Access SF Spreading Factor UL Uplink col, QoS, latenc y , communication range and deplo yment cost of each technology [16]. LPW AN application scenarios are categorized and some important parameters to be considered for each speciﬁc scenario are studied. Research challenges and recent technical adv ancements of each technology are not discussed in detail. Different from the above mentioned surve ys [9], [16], our surve y is focused on LoRa networks. W e study the recent performance measurements of LoRa networking [26], [32]– [36], [36]–[50] to understand and de vise a taxonomy for the research problems of LoRa networking. The recent so- lutions [21], [41], [51]–[76] are further discussed in detail to understand the adv ancements of LoRa technology . Finally , we present some open issues that could further improve the performance of LoRa networking. A surve y on LoRa technology has been recently pub- lished [77]. It discussed the literature, solutions and open 4 Fig. 1: LPW AN Netw ork architecture issues without any classiﬁcations. In our article, a clear taxon- omy has been devised. Based on this taxonomy , the challenges, current solutions and open issues are discussed with tabulated version of system analysis and hardware experiments. The tax- onomy provided in this article facilitates a clear understanding of the challenges, solutions and open issues. In addition, some of the recent contributions to LoRa networking [51], [52], [78], [79] are also discussed in our article. The rest of this paper is structured as follo ws. Section II giv es a brief description of LoRa technology . Section III lists the existing deployments of LoRa networks and their advantages. Section IV in vestig ates the research challenges of LoRa networking and de vises a taxonomy . Section V giv es a comprehensiv e study on ho w these research problems are tackled by recent solutions. Section VI discusses some open issues that still needs to be addressed and Section VII concludes the article. I I . A B R I E F T E C H N I C A L BA C K G RO U N D O F L O R A This section brieﬂy describes the technical features of LoRa. LoRa operates in unlicensed sub-GHz ISM band (900MHz in USA and 860MHz in Europe). Using 125KHz, 250KHz and 500KHz of bandwidth, smaller payloads of up to 250 bytes can be transmitted over a distance of 5-15 Km and the system can last up to 5-10 years consuming low power according to the recent report [10]. A LoRa system comprises of end-devices, gate ways, network and application servers. Figure 1 depicts the architecture of a typical LoRa system. end-devices collect information and send them to Gatew ays. Gatew ays relay messages between end-de vices and network servers. A network server is conﬁgured to direct messages to appropriate application servers for processing. There are three operating modes for LoRa. LoRa end-devices must implement Class A operating mode. Other optional modes like Class B and Class C can also be utilized. The end- devices operating in Class A and Class B modes are generally battery powered while the end-devices operating in Class C is mains powered. Class A utilizes less energy than Class B and C. In Class A, after sending conﬁrmed messages, end- devices expect an ackno wledgment (A CK) from the Network server during two pre-agreed time-slots kno wn as “recei ve Fig. 2: Class A receiv e windows windows (R W)”. Figure 2 depicts the R Ws of Class A op- erating mode. Frequency and data rate of the ﬁrst R W is the same as the uplink transmission parameters whereas the second slot operates on pre-agreed parameters to improv e the robustness of message transmissions. end-devices do not expect replies from the serv er for unconﬁrmed messages. Class B operating mode opens additional receiv e windows scheduled by gate ways through beacon packets. Class C mode has no downlink restrictions and can recei ve downlink messages any time whene ver it is not in a transmitting state. In general, LoRa denotes the physical layer while Lo- RaW AN denotes the MA C layer communications and netw ork- ing in LoRa stack. LoRa. The physical layer of LoRa technology uses Chirp Spread Spectrum (CSS). Chirps are the signals whose fre- quency v aries linearly with time within the av ailable band- width. This attribute makes the chirp signals resilient to noise, fading and interference. Every LoRaW AN packet starts with a preamble of ten chirps and six synchronization chirps followed by the data. Each chirp can modulate multiple chips (data bits). The number of data bits modulated depends on the parameter Spreading Factor (SF). For example, nine bits can be encoded in a chirp using SF9. A message sent with higher SF takes more time on air and reduces the data rate but improv es resilience to noise. LoRa modulation also has tw o other parameters namely Bandwidth and Code Rate (CR). The bandwidth can be set to 125KHz, 250KHz and 500KHz and the CR can be set to 4/5, 4/6, 4/7 or 4/8. LoRaW AN. The LoRa community often refers LoRaW AN as a “MA C in the cloud” design [52]. Gatew ays are the forwarders acting based on commands from the servers. All MA C decisions like data-rate, handling A CKs are decided at the servers. LoRaW AN MA C emplo ys tw o modes to di vide air-time between end-devices for handling collisions. The ﬁrst mode is the ALOHA MA C that allows end-devices to transmit as soon as they wakeup and exponential back-of f is applied in case of collisions. The second mode is the TDMA scheduler where the network server allots time-slot for each end-device to transmit their messages. A. Unique pr operties of LoRa LoRa technology has some unique properties making it a widely used technology . The unique properties are (i) Ultra- long distance, (ii) Low cost and complexity devices (iii) 5 Long lifetime of nodes, (iv) concurrent reception capacity of gate ways and (v) rob ustness in Doppler ef fect. All these unique properties are e xperimentally veriﬁed by [80]. Ultra-Long distance: In Line-Of-Sight (LoS) communi- cations, the longest SF12 can achiev e a distance of up to 9Km with Packet Reception Ratio (PRR) > 70% and the smallest SF7 can achiev e a distance of 5Km for PRR > 70%, according to the report in [34], [60]. In Non-Line-of- Sight (NLoS) scenarios comprising of b uildings, the longest distance achiev ed is around 2Km [79]. It is also noticable that the communication distance is affected by the parameters Bandwidth, SF , transmission po wer and coding rate [40]. Low cost and complexity: The LoRa devices are fabricated such that they are not complicated hence reducing the price. Reduced complexity also reduces the ov erheads incurred dur- ing communications. For example, a sophisticated CSMA is not employed instead a CAD is emplo yed that will just check for preambles in the channel before transmission. There is no signalling ov erhead like other traditional communication networks. Whenever a node wants to transmit, it wakes up, checks for channel status, transmits and goes back to sleep. Long lifetime: The LoRa consumes around 120-150 mW during transmission and 10-15 mW for MCU operations based on different radios and host-boards used. This can be extrapolated to 2-5 years in total life time while the duty cycle is v aried from 0.1% to 10% [80]. Concurrent reception of gateways: Current LoRa gate- ways are capable of concurrent reception on 8 channels. Even the same SF can be receiv ed on different channels. All the different spreading factors from SF7 - SF12 are orthogonal and transmissions with different SFs can be receiv ed on the same channel concurrently . Robustness to Doppler effect: Liando et al. [80] prove that LoRa signals are robust to Doppler effect. The CSS modulation used by LoRa is highly resistant to Doppler effect. Mobile LoRa end-de vices at a constant speed and in LoS can yield PRR > 85% [80]. I I I . E X I S T I N G D E P L O Y M E N T S O F L O R A N E T WO R K S This section explains the existing deployments of LoRa networking and their adv antages. There are man y use cases like building management system [22], smart agriculture [24], smart parking [81] and smart lighting [82]. some popularly known real-world deployments are summarized in T able III. An ov ervie w of these deplo yments is discussed belo w . Smart cities and Urban Deployments: Inﬂux of popu- lation toward cities demands a better way of governing and organizing amenities for optimal usage. Semtech’ s white paper [23] explains how LoRa LPW AN could provide efﬁcient usage and governance to make cities smart. Some applications that could improv e daily life of the people are Smart parking [81], Smart lighting [82]. These applications will impro ve people’ s living experience. Some in-ﬁeld deplo yments [23] are (i) w aste management in Seoul, North K orea, (ii) integrated sensing of Solar po wer plants in Carson city , Ne vada, USA, (iii) power monitoring in L yon and Grenoble, France. Seoul experiences humongous ﬂoating crowds e very day . As the crowd moves through the city dynamically , probing T ABLE III: In-ﬁeld deployments of LoRa netw orks LoRa Deployment Location W aste Management [23] Seoul, North Korea Solar power plant management [23] Nev ada, USA Power usage monitoring [23] L yon, France Power usage monitoring [23] Grenoble, France Smart meters [25] Gehrden, Germany Smart golf course [83] Calgary , Canada Smart Islands [84] Mallaorca, Spain the capacity of waste-bins became tedious. City management installed LoRa-enabled smart bins to periodically collect the capacity of waste-bins. This helped to clear the bins as soon as they are ﬁlled. This application giv es a 66% reduction in waste collection frequency , an 83% reduction in costs and a 46% increase in recycling. Carson City management in Nev ada found that effecti ve transition between legacy and solar power is important when solar efﬁcienc y reduces during cloudy climates and nights. LoRa-enabled monitoring system monitors the current en vi- ronment status of the solar-deployed site. Decision to use solar or leg acy power is based on this collected data. This system reduced 15% of operational expenses and boosted solar power output to 75,000 kWh of clean power because of proper transition between solar and legacy power . In L yon and Grenoble, power consumption monitors were deployed in households. This helped the residents to monitor their po wer usage and turn-off unwanted devices. This system helped to reduce power consumption by 16%. Smart meters: Semtech’ s white paper [25] describes and ev aluates the capacity of LoRa technology for smart metering applications. This system is deployed in Gehrden, Germany where the population is 15,000. Around 7000 households were installed with LoRa-enabled smart meters. This application helped to reduce the man-power utilized for monitoring power usage by transmitting meter-readings periodically to the gate- way . Smart golf course: Shaganappi point is a popular golf course in the city of Calgary , Canada serving a community of million people. Golfers play an average of 90,000 - 100,000 rounds of golf during April - November e very year . It is identi- ﬁed that slow play dev alues the ov erall experience. Improving this experience will help to retain customers. Hence, each golf cart is ﬁxed with a LoRaW AN sensor . With real time mov ement and location of golf carts, pause of play anomalies are detected and appropriate help is provided to speed up the play . Large coverage of golf course requiring periodic updates with low power consumption makes LoRa a perfect solution for this use case. Hence overall experience of customer is elev ated with maximization of rev enue [83]. Smart Islands: Mallorca is largest of Spain’ s ﬁv e Balearic Islands popular for white sand and turquoise water . Currently there are 25 people for ev ery meter of beach on the Island with 32% anticipated growth by 2030. Citizens have shifted their attitude to conserve natural resources in the Island. LoRaW AN sensors are installed to aid water management systems. This periodically reports water quality and lev els [84]. This system 6 has seen 25% water savings since installation. I V . T A X O N O M Y O F R E S E A R C H P RO B L E M S In this section, we study the research challenges of LoRa networking. The sev erity of these challenges is identiﬁed by in vestigating its effect on the operations of LoRa technology . Finally , a taxonomy is devised to categorize these challenges. Figure 3 illustrates the taxonomy of the challenges of LoRa networking. A. Ener gy Consumption The most important characteristic of LPW AN is its high energy ef ﬁciency . This becomes an important parameter to improv e the longevity of end-devices. LoRa networks are expected to work for a longer period of 5-10 years with minimal maintenance. Hence, power consumption becomes a major challenge for LoRa networking. end-device operations can be classiﬁed into (i) micro-controller operations and (ii) wireless transmissions. Po wer consumed for micro-controller operations vary according to the chosen host board but the power consumed for wireless transmissions completely de- pends on the LoRa technology . Charm [52] sho ws that wireless transmission extorts more power than micro-controller oper - ations. LoRa technology employs two techniques to reduce energy consumption, (i) consuming instantaneous bandwidth for transmitting a chirp signal and (ii) not employing heavy MA C protocols for scheduling. In spite of these techniques, end-devices consume more po wer than expected due to some unav oidable circumstances lik e retransmissions caused by channel impairments. B. Communication Range Large communication range is also an important rudiment of LoRa technology . Current LoRa technology relies on chirps spread spectrum, which is more resilient to interference. LoRa networking will be deployed in man y scenarios such as homes, hospitals, schools, forest, etc. end-devices will be placed in the locations open to air , closed by concrete or steel, etc. Aiming such div erse deployment conditions, signal attenuation, propa- gation losses and f ading ha ve to be countered to impro ve signal penetration thus improving the coverage of LoRa networks [35], [32]. It has been noted that gate ways can detect signals below a giv en threshold but cannot decode them. Devising a technique to decode these signals will improve communication range. Another important challenge is estimating the cov erage of LoRa networks. Chall et al. [79] study different models through empirical measurements in Lebanon. Demetri et al. [78] identiﬁed that LoRa’ s signal coverage is anisotropic. This is because LoRa signals trav el a longer distance and e xperience varying en vironments with dynamic and static obstacles in different directions. Mathematical models and systems for link quality estimation are still unexplored for LoRa. C. Multiple Access: LoRa networking aims to connect thousands of end-devices to the network, communicating over a conﬁned region and spectrum. Possibilities for these end-de vices transmitting data concurrently varies based on the application. Multiple access is to allow multiple end-devices to share the limited spectrum for communication. The multiple access problems inv olve two different aspects namely Link coordination and Resource allocation. Link coordination: Deploying thousands of devices require multiple-access to improve concurrent transmissions and avoid collisions. Links are coordinated through MA C protocols. The LoRa networking employs ALOHA and TDMA scheduler to coordinate links. These techniques cannot handle collisions while thousands of devices are connected to the network [42], [43]. Thus, new techniques for handling collisions and coordinating links are required. This will help to upscale the density of LoRa deployments [44], [45]. Resource Allocation: In LoRa technology , the transmission is controlled by the parameters Spreading Factor (SF), T rans- mission power (TP), Bandwidth (BW) and Channel. V arying these parameters result in different transmission qualities. This can be lev eraged to improve concurrent transmissions. Dynamically allocating reasonable resources to end-de vices based on the deployed en vironment improv es multiple access and thus scalability [85]. D. Err or Corr ection LoRa technology communicates data ov er long distances. While the message is transmitted over the air , it is possible for the data to get corrupted or lost due to channel ef fects, en vironmental conditions or collisions. Existing error correc- tion schemes of LoRa networking, like hamming code, cannot aid data corruption or loss efﬁciently [56]. LoRa technology offers different spreading factors to make the signal more resilient to noises. SF12 is the stronger spreading factor but it takes more time-on-air . Even these signals can also be corrupted in dense en vironments [33], [34]. There are two types of current solutions namely (i) channel coding and (ii) interference cancellation. Channel coding . A recent channel coding technique pro- posed is DaRe [56]. DaRe is an application layer coding to retriev e lost data using redundant data. Sandell et al. [59] explain that this technique cannot aid bursty packet loss and has some limitations that bounds the performance. Interference Cancellation: Even though channel coding aids error , they cannot guarantee error correction in the case of collided signals. Interference Cancellation will extend error correction by untangling and extracting data from collided signals. Recently proposed technique Choir [51] and Netscatter [55] cancels interference but Choir’ s [51] limitation is pointed out by Netscatter [55]. Netscatter [55] can cancel collisions of 256 concurrent transmissions, which is not adequate to handle thousands of end-devices by a single gate way . For example, a large-scale temperature monitoring system requires all end-devices to transmit data at the same time [25]. Hence, Interference cancellation is still an open problem that should be addressed to improv e the performance of LoRa Networking. 7 Fig. 3: T axonomy of LoRa Research Challenges E. Security For all computer communications, security is a major con- cern. There are many security attacks like eavesdropping, selectiv e forwarding and node impersonation [86]. All the abov e mentioned attacks try to obtain the key used for encryption. If this key is compromised, the entire system can be brok en. Currently , LoRa technology uses a symmetric key cryptographic technique with AES-128 bit encryption. Existing LoRa technology generates the key and never updates it. Hence, K ey generation and Ke y update mechanism is a major concern. Thir d-party authorization is required when the application and network service providers are dif ferent. So, these applications require third-party authorization to ensure priv acy and security [47], [75], [76]. V . C U R R E N T S O L U T I O N S In this section, we summarize the recent measurements and current solutions proposed to address the problem power consumption, communication range, error correction, multiple access and security . Existing works can be classiﬁed into two categories as performance measurements and current solutions tackling the abov e challenges. All the recent experiments, measurements and simulations use Class A end-devices unless speciﬁed. T able IV summarizes performance measurements and case studies conducted on LoRa. T able IV differentiates the mea- surements made through theoretical analysis, test-bed and simulated ev aluations. It is to be noted that no performance measurements hav e been made on power consumption and error correction capabilities. T able V enumerates recent solu- tions to improve the performance of LoRa networking. The current solutions addressing more than one problem can also be identiﬁed in T able V. A. Ener gy Consumption V arious techniques are employed to improv e battery lifetime of the end-devices. Some works propose techniques (i) to harvest ambient energy from the environment [87], [54], [53]; (ii) to use backscatter signals for transmission, [54], [53], and (iii) to detect and decode weak signals and increase data rate to reduce po wer consumption [52]. LoRa Backscatter [54] proposes a backscatter system for LoRa based on CSS modulation. Data can be transmitted up to 2.8 Km while consuming only 9.25 µ watts of power at the rate of 37.5 Kbps. Power consumption is reduced by nearly 1000 × than standard LoRa technology . These passive RF chips can be powered through solar panels attached to them. This technique is also analyzed ov er home/ofﬁce sensing, precision- sensing of agriculture devices and epidermal de vices to prov e their ef ﬁciency . PLoRa [53] proposes a hardware and software co-design to enable battery-free LoRa networks, operating on the en- ergy harvested from solar devices. The proposed PLoRa tag transmits data by backscattering ambient LoRa transmissions without external excitation signals unlike LoRea [88] and LoRa Backscatter [54] which uses dedicated hardware for generating excitation signals. The acti ve LoRa signal emitted by a gate way or a node is con verted into Passiv e LoRa signals to send data using ON-OFF keying technique. The power consumption of PLoRa is 250 × smaller than the standard LoRa technology . Charm [52] improv es the battery lifetime up to 4x the stan- dard LoRa technology by avoiding retransmission of the weak signals. This technique is discussed in subsection V -B. F ADR [62] reduces the power consumption of standard LoRaW AN by 22%. W ireless po wer char ging has been a promising solution to handle energy consumption problem for wireless sensor networks [89]. Realizing them on low-cost LoRa hardware has been done in [90]. A circuit has been designed to enable wireless po wer transfer on LoRa enabled sensor nodes. Gao et al. [91] in vestigated energy fairness problems in LoRa networks. Due to large differences between data rates used by dif ferent end-devices, end-devices far a way from the gate way have to use a low data rate and spend more ener gy to transmit certain amount of data. T o make energy consumption more fair across all end-de vices in a LoRa network, Gao et al. [91] propose to deploy more gate ways to allo w end-devices to use high data rates to reach at least one gateway . T o make the network transmission more efﬁcient, a network model is dev eloped and used to allocate network resource to each end- device. A heuristic search based network resource allocation algorithm is de veloped to ﬁnd the best netw ork setting for each end-devices. Key Insights. Ambient energy harvesting is used to make LoRa de vices battery free [53]. One of the most power con- suming operation in LoRa wireless transmissions is generating carrier signals [52]. One way to reduce this part of power 8 T ABLE IV: Performance measurements of LoRa Articles Communication Range Packet Delivery Multiplexing Security T/M S T T/M S T T/M S T T/M S T Harris et al. [35] Fuidiak et al. [32] Petajajarvi et al. [33] Petajajarvi et al. [34] Haxhibeqiri et al. [36] Mikhaylov et al. [37] Magrin et al. [26] Semtech White paper [38] Bankov et al. [39] Haxhibeqiri et al. [36] Angrisani et al. [40] Blenn et al. [41] Lauridsen et al. [42] V ejlgaard et al. [43] Zhu et al. [44] Orfanidis et al. [45] Ferre et al. [46] Aras et al. [47] Butun et al. [48] Miller et al. [49] Oniga et al. [50] Liando et al. [80] T able notes T/M - Theoretical / Mathematical Analysis S - Simulated ev aluation T - T estbed ev aluation consumption is to utilize the backscatter signals [54]. Another method is to lev erage passiv e chips [92] for carrier generation. Charm [52] identiﬁes that LoRa is able to receive weak signals but not able to decode them. W ireless power transfer [89] is not feasible owing to the complex, costly hardware extensions in low power , low cost LoRa modules. Placing more gateways and dynamic allocation of TP addresses energy consumption problem and impro ves netw ork lifetime. B. Communication Range T estbed Measurements. Semtech’ s white paper [38] ev alu- ates the capacity of LoRaW AN in dense urban environments. T en gateways operating on 8 channels were used for trials with 100 sensors as end-devices to transmit at different data rates to mimic like 10,000 end-devices. Three phases of experiments with varying quantity of packet transmission were conducted to ev aluate packet deliv ery . The ﬁrst phase contains 250,000 volume of packets being transmitted at the rate of 104 packets per hour . The second phase generates 500,000 packets at the rate of 209 packets per hour and the third phase generates 1,000,000 packets at the rate of 417 packets per hour . For all the three phases, the achieved deliv ery rate is more than 95%. This paper notes that multiple gateways will scale-up the network and improv e communication range as end-devices can communicate with more than one gatew ay . Nav arro et al. [89] and Haxhibeqiri et al. [36] e valuated communication range of LoRa in Industrial en vironments. The industrial en vironment at Royal Flora Auction Center , Netherlands, cov ering 250000 m 2 of both indoor and outdoor spaces. LORANK [90] Gateway was ﬁxed 6 m above the ﬂoor and W iMOD-IM880A [93] end-devices were attached to the trolleys 1.7 m above the ground. The nodes were triggered to transmit at SF7 and SF12. Measurements were taken from 43 measuring points covering indoor , outdoor spaces. Fifty packets were sent from each test point. Measurements show that a maximum of 6000 end-de vices can be handled by a single gatew ay with a packet loss rate less than 10%. Packet loss is around 6% when less than 3500 end-devices are used. Petajajarvi et al. [33] analyse the range of LoRaW AN with 14 dBm T ransmit Power (TP) and the largest SF at Oulu, 9 T ABLE V: Summary of recent solutions for LoRa Challenges Article Multiplexing Power Consumption Communication Range Error Correction Security Choir [51] Charm [52] LoRa Backscatter [54] PLoRa [53] Netscatter [55] DaRe [56] Blenn et al. [41] Bor et al. [21] Chall et al. [79] Demetri et al. [78] Georgiou et al. [57] Donmez et al. [58] Sandell et al. [59] Haxhibeqiri et al. [60] Reynders et al. [61] Abdeel et al. [62] Reynders et al. [63] Pop et al. [64] Cuomo et al. [65] V an et al. [66] Mikhaylov et al. [67] Cattani et al. [68] V oigt et al. [69] Lee et al. [70] Kim et al. [71] Na et al. [72] Girard [73] Kim et al. [74] Naoui et al. [75] Liando et al. [80] Finland. The experiment is conducted for 14 days during spring and summer . The population of Oulu is around 200,000 people with high rise buildings. Throughout the experiment, Kerlinks LoRa gatew ay [94] is ﬁxed on a tower 24 m abov e the sea le vel with -137 dBm sensitivity in order to ﬁnd the maximum communication range. Semtech 1272 transcei ver [95] is used as an end-device. For on-ground measurements, end-devices are ﬁxed on a car’ s roof-rack, approximately 2 m abov e the ground, which drove around major cities at 40 Km/h - 100 Km/h. F or on-water measurements, end-devices are ﬁxed on a radio mast of the boat. end-devices send packets periodically including GPS coordinates. The SF of end-de vices is set to SF12 because the goal was to ﬁnd the highest possible cov erage. Maximum range noted on ground and water is 15 Km and 30 Km respectively . T otal packet loss ratio for on- ground measurements and on-water measurements become 34% and 32% respectively . With these measured data, channel attenuation model is calculated for areas similar to Oulu. Research solutions. Du et al. [96], [97] proposed a solution to improve the communication range of sparse wireless sensor networks. Choir [51] identiﬁes an intrinsic property of LoRa radios in which the carrier frequency varies by a small bound (902.4 MHz instead of 902.7 MHz) because of cheap radios. This is exploited to disentangle collided signals and extend the range up to 2.64 × the standard LoRa technology . The nodes that are far away from the gate way transmit signals 10 whose SNR goes belo w the noise ﬂoor . It is assumed that the neighbouring nodes send data that do not vary to a greater extent. These physically co-located nodes, far away from the gate way , are coarsely synchronized through Class B beacons to transmit data at the same time to enable constructiv e interference thus improving the SNR abov e the noise ﬂoor . Gatew ays can receive and decode this collided signal to obtain approximate data of the region far a way from gateways. This technique improves the communication range by 2.64 × the standard LoRa technology . Charm [52] proposes a new hardware and software co- design to extend cov erage and battery life of LoRa devices. This is achiev ed by allowing multiple gate ways to send weak signals (that cannot be decoded by a single gatew ay) to cloud and coherently combine them to decode data. Programmable auxiliary hardware attached to the gate way improves the gate way’ s ability to detect very weak signals which cannot be directly detected by gatew ays. Joint decoding algorithm uses a heuristic approach to select signals to be combined at the cloud. Results show that the range is impro ved 3 × and battery life is impro ved 4 × the standard LoRa technology . LoRa Backscatter [54] discussed in subsection V -A can send data to a recei ver located 2.8Km aw ay . Sensor Networks Over Whitespaces (SNO W) [98] was ﬁrst designed for sensor networks to be connected over a wide area. As the traditional sensor networks cannot communicate to a longer distance, TV whitepsaces that could communicate a long distance is exploited. Scalability and energy efﬁciency is achiev ed by splitting carrier into several sub-carriers with parallel packet receptions. The PHY layer handles OFDM modulation and the MA C layer handles sub-carrier allocation. This technique is extended by Rahman and Saifullah [99] to integrate multiple LPW ANs and improve the communication range speciﬁcally in infrastructure-restricted rural areas. The nodes located far away communicate with gate ways ov er white spaces meant for TV signal communication. This technique is implemented on generic GNU radios. Its function on LoRa devices is still ambiguous. Key Insights. The abov e measurement experiments throw light on (i) placing more gateways to improv e network density , cov erage and reduce energy consumption (ii) less packet loss in netw orks with sparse end-de vice placements [33]. Besides, we can infer the following insights: • Radio imperfections occur in LoRa due to cheap radios put into use. • These radios generate slightly dif ferent carrier frequencies than speciﬁed [51]. • LoRa g atew ays can receive weak signals but cannot decode them [52]. • Extending cov erage of LPW AN’ s is experimented with GNU radios b ut its operation on commercial LoRa de- vices has not yet been studied [99]. C. Err or Corr ection In this subsection, the research solutions for error correction are classiﬁed into channel coding and interference cancella- tion. The solutions are explained in detail before concluding with ke y insights. 1) Channel Coding: A ne w application layer data recov ery technique called DaRe [56] is proposed based on Con volu- tional and Fountain codes. This technique extends data with redundant information. These redundant data are chosen from the previous data units so that the lost frame can be calculated from the other received frames. The disadvantage is that previous data units should be buffered in the memory for computing redundant information. This makes the generator matrix banded thus inducing difﬁculty to create degree of dis- tribution according to L T codes. Sandell et. al [59] sho w that the memory affects performance and complexity . While DaRe uses a complex Gaussian elimination making the decoding process complex, an optimized decoding technique is proposed in [59]. Sandell et al. [59] analyses the technique proposed in [56] and shows that for lar ger packet loss probabilities, reducing the code rate increases the interference leading to reduced efﬁcienc y of data recov ery . Showing the relationship between latency of decoding algorithm and packet loss probability , a less complex decoding algorithm like Accumulati ve Gaussian Elimination is proposed by Du et. al [97] to reduce latency . Finally , the paper concludes that data reco very through redun- dancy techniques increase number of transmissions and thus collision. So it cannot be used solely to aid packet loss. 2) Interfer ence Cancellation: Choir [51] and FTrack [100] propose nov el solutions to disentangle and decode collided signals. Using constructi ve interference, a sparse o vervie w of the data is obtained from a group of geographically co-located end-devices far away from the gateway . T est bed e valuation shows that throughput is improved 6.84 × than the standard LoRa. It is also pro ved that Choir yields better results than multiple antenna deplo yments. NetScatter [55] improv es interference cancellation. It the- oretically proves that Choir [51] can only decode 5-10 concurrently transmitting devices. A ne w distributed coding technique based on CSS is proposed to decode concurrent transmissions below noise ﬂoor in a single FFT operation. Experimental results show that this technique can decode 256 concurrent transmissions with 14-62 × improvement in throughput and 15-67 × improvement in latency when com- pared with e xisting interference cancellation techniques. Key Insights. The key insights regarding error correction are listed as follows. • An intrinsic property of LoRa to deviate from carrier fre- quency is identiﬁed and e xploited in Choir [51] to correct errors because of signal collision. Choir is analyzed and its inability to scale-up is identiﬁed and a ne w distrib uted error correction technique is proposed in Netscatter [55] for scalability . • Recently proposed channel coding technique, DaRe [56], and its analysis [59] sho ws that there is a heavy potential for characterizing ef ﬁcient channel coding techniques for LoRa to improve error correction, thus improving reception rate and network lifetime. D. Multiple Access In this subsection, the solutions for multiple access are divided into Link coordination and Resource allocation. The 11 measurements and solutions for each category are explained below with their key insights summarized at the end of each category . 1) Link Coor dination: Measurements. Bankov et. al [39] identiﬁes four important issues. The ﬁrst issue is whether the gate way should listen to the channel during the interval, T1, between frame reception and transmission of response. There arises a problem if the channel is busy with other scheduled transmissions in that speciﬁc interv al T1. This might cause delay to the A CKs, leading to unwanted retransmissions. The proposed solution is to cancel the pending transmission that may cause collisions at the end-device and transmit A CKs on the downlink channel. The second issue occurs when two transmitted frames are ov erlapping in the same time interval over different channels. Gate way would not be able to acknowledge both messages at the speciﬁed window with single downlink. The third issue is the limited interval for retransmission. Due to the abo ve said factors, A CKs may take more time to reach an end-device which will increase the retransmission probability ev en for a successfully deliv ered message. The authors recommend increasing the delay interval or use exponential back off to counter this issue. The fourth issue arises when there is no optimal polic y to select the data rate for downlink. Simulations conducted show that pack et err or rate and pack et loss ratio increase with traf ﬁc due to improper link coordination. Research Solutions. Reynders et al. [61] address Lo- RaW AN’ s scalability and reliability through a novel MA C protocol, RS-LoRa. This technique works in two phases. The gate way sends coarse-grained information of allo wed TP and SF for each channel as Class B beacons in the ﬁrst phase. In the second phase, each end-device selects one parameter combination from the beacon that better suits the node. As- signing different SFs with dif ferent parameter combinations helps to alle viate the Capture Ef fect. Reliability of network performance is improved by decreasing the Packet Error Ratio up to 20% than the standard LoRaW AN. This technique proves its superiority to the standard LoRaW AN through NS-3. Based on measurements obtained through real-world ex- periments, Haxhibeqiri et al. [60] build a simulation model to study the scalability of a single cell LoRaW AN based on interference. It is showed that LoRa physical layer is robust and can send six times trafﬁc than the pure Aloha with 125 KHz bandwidth. Based on end-device density and their data rates giv en in [101], simulations are carried out to determine node density for different IoT applications. A series of experiments are conducted to identify the po- tential of Channel Activity Detection (CAD) and Ideal-CSMA on a dense network of 50 nodes by Liando et al. [80]. The results sho w that Ideal-CSMA f ails to provide high reception rate when nodes synchronously perform channel detection. Hence, the authors devise a CSMA-CAD with four additional preambles in the packet to achiev e doubled PRR. Whenev er a preamble is detected, the transmitting SF is randomized and the channel is sensed again for transmission. The CSMA and CSMA-x are simulated on NS-3. CSMA- x works similar to CSMA but it senses the channel for the gap of x ms. The results of this model are compared with the outcomes of real-word test-bed and the results from other works to prove the model’ s accurac y . It can be seen that e ven though CSMA cannot provide better results for dense nodes, CSMA-10 is achieving better performance than CSMA. The performance comparison of p-CSMA and CSMA is conducted by Kouv elas et al. [102] for small scale networks expound that p-CSMA is an important step to be taken for improving the scalability of LoRa networks but it has not been realized yet on real-world devices. There are many retransmission policies devised for wireless networks [103]–[106] and wireless sensor networks [107]– [109]. A joint retransmission scheme with compression and channel coding is dev eloped for single-hop networks with energy constraints lik e LoRa [110]. Its energy efﬁcienc y has been theoretically ev aluated. This retransmission scheme retransmits the last q failed data blocks along with a new data block using compression and coding schemes. Although the performance has been prov ed theoretically , its implementation with the required computational resource on real-time testbeds will be more helpful to understand the practical efﬁcienc y for real-world deplo yments. Key Insights. T ailoring different transmission parameter combinations for different end-devices while considering net- work density is prov en to be capable of impro ving link coor- dination [61]. Liando et al. [80] lev erage the fact that LoRa gate ways can receiv e packets with dif ferent SFs and the same frequency simultaneously to reveal that PRR can be doubled if the SF is randomized on detecting a preamble during CSMA. It is theoretically prov en that a retransmission scheme employing data compression and channel coding improves link coordination. 2) Resour ce Allocation: Measurements. Semtech’ s White paper [25] e valuates the capacity of LoRa technology for smart metering applications. This system is deployed in Gehrden, Germany . Around 7000 households were installed with LoRa- enabled smart meters. 11 K erlink-V1 gatew ays [94] were mounted on rooftops with 30cm/70cm half wav e dipole an- tenna. A simple meter protocol was employed for reading a sev en-digit register . The payload allocates one byte for status and three bytes for each re gister . The do wnlink payload length is ﬁxed as 10 bytes. Meters are conﬁgured to send unconﬁrmed payload e very 15 minutes and conﬁrmed payload once a day . Class-C end-devices are utilized for this experiment. 24-hour raw data is used for the experiments. It is sho wn that a gate way with average throughput can handle 470,000 messages per day . It is demonstrated that the network can be scaled up locally by adding gate ways. It is also sho wn that the ADR algorithm improves the network capacity by adjusting the data rate, frame repetition rate and channel allocation. Petajajarvi et al. [34] conducted experiments in the Uni- versity of Oulu, Finland. LoRaMote [111] is used as end- device for measurements. These end-devices were conﬁgured to send messages to the base station e very 5 seconds with no A CKs, no retransmissions and no ADR. End-devices are conﬁgured to transmit at six different channels. The gateway is the same as used in [33]. The packet delivery was above 96.7% and 95% when the end-device was static and mobile respectiv ely . Similar results ha ve been observ ed in [38] and 12 [26]. Measurements show that most of the campus is cov ered by SF7 itself. Interesting results were obtained while v arying physical parameters. Farthest location is not reached by SF7 and BW125. Howe ver , 60% of the packets were correctly receiv ed from the same point with SF7 and BW250. For po wer consumption ev aluations, RN2483 based LoRa module was added to a sensor and actuator kit with a K eysight’ s power analyser . It is noted that energy consumption of the same packet transmission varies by more than 50% between the maximum and minimum v alues. This measurement stresses on the usage of ADR to reduce ener gy consumption. Angrisani et al. [40] assess the performance of LoRa under critical noise conditions. Transmitter , powered through a power bank, placed 10m away from the receiver . A White Gaussian noise is generated to corrupt the transmitted signal. The ﬁxed parameters distance, payload and preamble are set to 10m, 1 byte and 8 symbols respectiv ely . On varying SF , BW and CR to all possibilities, authors claim that an increase in BW increases the packet loss with lower SFs. But packet loss is decreased with larger SFs. The authors also state that an increase in CR can trade off an increase in BW and SF . Finally , it is concluded that LoRa is highly robust to high noise lev els and recommend further in vestigation by varying the ﬁxed parameters of this experiment. Iov a et al. [112] in vestigate the performance of LoRa in mountain regions and identify the factors affecting transmission parameters. Hakkenber g et al. [113] and Neumann et al. [114] ev aluate the performance of LoRa in both indoor and outdoor en vironments and recommend that transmission parameters hav e to be varied according to the deployed en vironment. Describing the operations of LoRaW AN, Augustin et al. [115] ev aluate LoRa’ s receiver sensitivity , network coverage using Freescale KRDM-KL25Z development board [116] with Semtech 1276 transceiv er [95] as end-device and Cisco 910 industrial router [117] as gatew ays. Gatew ay is connected to The Things Network server to monitor recei ved packets. Gatew ay is placed indoors and end-devices were kept moving outdoors in the urban en vironment. Transmit power of the end-devices was set to minimum 2dBm with a 3-dBi antenna. Packet losses start at 100m. The measured RSSI values were slightly above the speciﬁed values for each SF . For network cov erage experiments, the gate way was placed 5 m abov e the ground level and end-devices were kept in a car with default transmit power 14dBm speciﬁed in [118]. PRR is tested for SF7, SF9, SF12 at various distances with ACK and retransmission turned off. At 2800m, SF7 achiev ed 0% PRR while SF12 deliv ers about 80% of the packet. The authors ﬁnd that communication coverage is directly proportional to SF v alues. Blenn et al. [41] analyse the 9.4GB data obtained from the The Things Network during December 2015 and July 2016 from 1618 unique devices. It is inferred that 3.7% of unique packets were recei ved by two g atew ays, 1.1% of unique packets were recei ved by three gate ways. A verage payload size is 18 Bytes where 93.7% of captured payloads are less than 50 bytes and 50% of the payloads are less than 19 bytes. It is observed that using higher SF and higher transmission power results in lo w packet loss. Cattani et al. [68] conduct experiments to understand the effect of tuning PHY parameters and en vironmental factors on LoRaW AN communication reliability and energy efﬁcienc y . Experimental results sho w that, for end-devices far away from the gatew ay , Packet Reception Ratio of fastest PHY setting is only 10% lower than the slowest setting. Hence, the authors recommend selecting high data rate and high transmission power for the end-devices far away from the gate way . On studying the effect of environmental factors, it is shown that signal strength is decreased by 6dBm at 60 ◦ C. Even this small deviation can increase packet loss in the messages transmitted by end-de vices far away from the gateway . Mikhaylov et al. [67] study LoRaW AN’ s susceptibility to inter-netw ork interference. Experiments with and without an interferer between transmitter and receiver giv es an insight to design a protocol for ﬁnding dynamic communication parameters. Experimental result shows that a personalized communication parameter for each end-device will aid scala- bility . Research Solutions. Fair Adaptive Datarate Algorithm (F ADR) [62] is proposed to select SFs and transmission power to achiev e data extraction rate among all end-de vices. SF is allocated based on the method described in [63], using RSSI and po wer lev els. End-devices are grouped based on re gions. This technique simulated in LoRaSim achiev es 300% higher fairness than the technique proposed by Bor et al. [21] and 22% higher fairness than the technique proposed by Reynders et al. [63] while reducing network energy consumption by 22%. Bor et al. [21] consider bandwidth and transmission po wer for scalability analysis through simulation. Georgiou et al. [57] dev elop a mathematical model for a LoRaW AN netw ork with a single gatew ay by also considering other unique LoRaW AN features like modulation and radio duty-c ycling. A mathemati- cal in vestigation on link outage, considering signal below SNR threshold and capture effect, is carried out in order to study their effects on scalability . It is inferred that the latter reduces network performance with increase in density of end-de vices, which hinders the scalability . Different from [21], which studies the scalability of Lo- RaW AN network through LoRaSim, Mikhaylov et al. [37] present mathematical analysis without considering many fac- tors. Bor et al. [21] build a simulator ruminating Bandwidth and T ransmission Po wer for modeling uplink beha vior . Three experiments are conducted. The ﬁrst experiment with a single gate way and multiple end-de vices with homogeneous commu- nication parameter infers that a single gate way can support 120 end-devices per 3.8 hectares. The second experiment contains a single gateway and heterogeneous communication parameters, such that the end-device’ s uplink air time is decreased, showing 13 × increase in node density than the pre- vious experiment. The ﬁnal e xperiment with multiple gate ways improv es data extraction rate. T wo suggested guidelines are to dev elop a protocol to decide communication parameters dynamically and to ev aluate optimal gate way placement for better scalability . Chall et al. [79] collect empirical data in Lebanon to verify various radio propagation models like Okumura-Hata [119], 13 [120], Cost-231 Hata [121] to ﬁnd their drawbacks and ﬁx them with additional proposals to make it acceptable for LoRa networks. Bor et al. [21] also propose a mathematical model for LoRa communication coverage based on the empirical data obtained over 2.6 Km of rural area and 100 m of built- up en vironment. Demetri et al. [78] compare link attenuation of LoRa signals in free space and Bor’ s model. It is shown that free-space model underestimates signal attenuation while Bor’ s model ov erestimates it. It is also claimed that Bor’ s model [21] need on-site measurements which is hard due to that most of the cov ered re gions comprise of transitional links, which are deﬁned as links with dynamic temporal link qualities. Hence, a new automated link quality estimation system without requiring on-site measurements is developed by Demetri et al. [78]. T o achiev e this, remote sensing spectral images from an open-source satellite is used. These images are fed as input to Support V ector Machine (SVM) [122] to classify dif ferent constitution of land coverage like water bodies, forests and buildings. Okumura-Hata model [119], [120] is modiﬁed to support LoRa link estimation on dif ferent land coverage. The results of SVM classiﬁcation is used to automatically choose and conﬁgure parameters of link estimation model. Margin et al. [26] implement a new NS-3 module to simulate dense urban environments. The link performance and measurements, signal attenuation due to buildings and other factors are giv en. Spreading Factor assignment is done based on power lev els of the end-device at the gate way . The gate way will bind with the end-de vice transmitting on highest receiv ed power le vel. 17 Gate ways are placed in a hexagonal grid around the central gateway , cov ering a 7.5 Km radius. T otally 10 4 end-devices are placed randomly . The e xperimental results show that densifying gatew ays such that each gate way cov ers 1200m can achieve a pack et deli very o ver 90%. But this increases collision as the number of end-devices using SF7 increases. The authors recommend that ADR mechanisms should be le veraged to counter such collisions. T wo techniques are proposed by Cumo et. al [65] to allocate SFs to end-devices. First technique EXPLoRa-SF allocates SFs based on RSSI of the end-device receiv ed by the gateway . EXPLoRa-A T guarantees T ime-on-Air equalization for all end-devices in addition to SF allocation. This is achiev ed by or dered waterﬁlling technique to e venly distrib ute channel load among end-devices in the network. Simulation of EXPLoRa- SF and EXPLoRa-A T in ” LoRaSim ” performs better than the basic ADR. Bor et al. [123] study the impact of transmission param- eter selection on communication performance and propose an algorithm to quickly identify the optimal transmission parameters for ener gy efﬁcienc y and reliable communication. It is shown that in vesting high energy for higher SF values does not always improve communication performance. From experimental results, authors claim that it is also possible to achiev e minimum energy efﬁciency while selecting desired transmission parameters based on application requirements. The proposed probing algorithm ﬁnds a transmission param- eter that halves the transmission power with PRR lar ger than a threshold. If not found, other settings that uses at most half the transmission power are probed. If a potential setting is not found, the algorithm employs an iteration bound to try other settings. It is shown that the proposed probing algorithm ﬁnds an optimal setting that uses only 44% more energy than the ideal setting within 285 probes. V an et al. [66] conduct experiment with single, multiple gate ways and various SF to study the scalability of LoRaW AN in NS-3 simulator . Error model combined with NS-3 Lo- RaW AN protocol is constructed through e xtensiv e baseband bit error rate simulations and used as an interference model. Experimental results show that usage of ACKs se verely affects uplink trafﬁc and ha ving multiple gatew ays impro ves scalabil- ity to a smaller extent. It is also sho wed that assigning dynamic communication parameters will help to up-scale node density . Reynders et al. [63] ﬁnd an optimal SF setting to reduce collision probability and distribute SFs and transmission pow- ers to decrease Packet Error Ratio (PER) of end-devices far away from the gate way . A routine to assign SFs and power control is dev eloped based on genetic algorithm. The key idea of this algorithm is to assign dif ferent SFs and power control to different nodes such that signals do not interfere with each other . Simulation of this technique in NS-3 sho ws that the PER of the overall network is reduced by 42% and the packet error ratio of end-de vices far a way from the gatew ay is reduced by 50%. Pop et. al [64] extends LoRaSim [21] by adding more features like A CK, downlink data messages and presents a ne w simulator called LoRaW ANSim . Same experimental settings used in [21] are used with additional downlink trafﬁc to study the scalability with downlink A CKs. It is inferred that scalability is hampered as handling ACKs reduce network performance. V oigt et al. [69] compare the usage of directional antenna and multiple gate ways to alle viate interference that arises due to dense deployments. It is shown that the gain of multiple gate ways outperforms the usage of directional antennae. Key Insights. The key insights regarding error correction are listed as follows. • Communications with higher SF and higher TP can reach a longer distance [41], [115]. • Network density can be scaled-up by adding more gate- ways and personalizing transmission parameters like data rate and channel allocation using ADR for each end- device [25], [66], [67], [69]. • Energy consumption v aries up to 50% while using lowest and highest transmission power to transmit the same packet. This sho ws that the lowest possible transmission power should be used for sa ving maximum energy . • It is not always required to increase SF if some locations cannot reach g atew ay . Increasing bandwidth on the same SF also increases PRR [34]. But this is not always the case at lower SF [40]. In vesting more energy by using high SF does not necessarily improv e communication performance [123]. Hence, v arying transmission parame- ters based on indoor/outdoor deployed en vironment gi ves better performance [113], [114]. • Only 10% difference is identiﬁed in PRR between fastest and lowest SF and TP settings for the end-devices located 14 at the farthest reachable point from gatew ays. Hence, lower SF and TP settings can help reduce resource consumption [68]. • Besides, the communication performance can also be improv ed by reducing unconﬁrmed messages [63], [66]. E. LoRa Security Measurements. On describing the LoRa network stack, Arsas et al. [47] explore the vulnerabilities of LoRa. This paper expounds four possible techniques to compromise the LoRa network. Firstly , compromising security ke ys. This is easier if an attacker can gain physical access to an end- device. Feasibility of this attack is demonstrated through experiments. Extracting security k eys from any end-device will enable the attacker to decrypt any message in the network. Secondly , The Jamming attack. The Jamming attack is an attack in which the communication channel is jammed with an intentional interference to corrupt the data signal sent in that channel. Reynders et al. [124] show that LoRa is also prone to Jamming attacks e ven if the chirp spread spectrum is robust to interferences. Demonstrating this through experiments, it is shown that a speciﬁc node can also be tar geted. Thirdly , The replay attack. It is an attack in which the intruder intercepts the message and resends it to the receiv er whenever the intruder wants to accomplish a particular task. Miller et al. [49] show that the consequence of this attack depends on the application scenarios. Finally , the wormhole attack captures a packet from a non-malicious node and this nev er reaches the server . Some credentials stored in the packet is v alid and can be used at any time in the network. Butun et al. [48] surveys and veriﬁes the feasible security threats of LoRa V1.1 with Scyther security veriﬁcation tool [125]. The attacks veriﬁed are (i) RF Jamming attack, (ii) Replay attack, (iii) Class B beacon synchronization attack, (iv) Network traf ﬁc analysis, (v) Man-In-The-Middle attack. Donmez et al. [58] identify the security vulnerabilities of V1.1 in the backward compatibility scenario. The security vulnerabilities of LoRaW AN V1.0.2 and corresponding so- lutions added in LoRaW AN speciﬁcation V1.1 to mitigate them are discussed. The open vulnerabilities during backward compatibility scenarios are discussed and countermeasures are proposed to mitigate them. While the speciﬁcation is explain- ing only one backward compatibility scenario, this article veriﬁes all possible scenarios to ﬁnd other vulnerabilities. Research Solutions. The discussion on proxy-based key establishment for securing messages in the IoT context in [126], [127], [128], giv es an insight on the proxy-based key establishment for LoRa. Naoui et al. [75] discuss the possibilities of applying proxy-based key exchange systems to enhance LoRa security . Bit ﬂipping attack, an attack by which the bits of ciphertext is changed, is countered in [70] using circular shift and swap techniques. K e y Generation. T omasin et al. [76] analyse the security of join procedures, especially through On-The-Air-Acti vation (O T AA) of LoRa. Join procedures are acti vated at least once- a-day to check whether the node is still connected to the Network. DevNonce is a random number in the Join-r equest message. DevNonce should be unique for each Join Request . Network Server declines or excludes the node sending Join- r equest with a previously used DevNonce . This paper identiﬁes the probability of regenerating a used DevNonce and the scenario where malicious node ﬂoods Network Server to register possible DevNonce’ s randomly , making the future join request of non-malicious nodes tougher . A random number generator algorithm is proposed and the size of DevNonce is increased from 16 to 24 bits to ov ercome these shortcomings. Kim et al. [71] focus on resolving the following three problems. The ﬁrst problem is the current DevNonce system. Join-r equest sent by benign end-device can be mistaken as a replay attack by network server because an end-device can regenerate old DevNonce . Not storing all the past DevNonce’ s will not prev ent replay attack. The second problem is the 24- bit DevNonce proposed in the article [76], that is incompatible with the current LoRaW AN speciﬁcation. The third problem is the token-based scheme proposed in [72]. This prev ents replay attack effecti vely but does not consider the scenario where the token is lost. For example, when the end-device reboots, the token is lost. This paper proposes two types of Join-r equests called Initial and Non-Initial Join requests. Non-Initial Join- r equest uses the token and it changes after each join procedure is complete. Current De vNonce may be regenerated. As the Ini- tial Join-r equest rarely occurs, regeneration of old DevNonce is negligible. When the token is lost, the node reboots and initiates Initial Join-r equest . These techniques are proved to enhance security through theory and experiments. Oniga et. al [50] explicates different security aspects of LoRa and proposes a secure network architecture framework. Implementation of this model under dif ferent testing scenarios recommend techniques for better data security and priv acy of LoRa based applications. Thir d-party authorisation. Girard [73] pointed out that both application and network session keys being generated at the network server will create a conﬂict of interest between network and application service providers. The network server and application server can deri ve both Application and Net- work session ke ys which is not secure if two different or gani- zations are in volv ed. So, a trusted third-party key management architecture is proposed. The problem of Ke y management and update mechanism is well addressed by the techniques proposed in [129], [130] for wireless sensor networks. As Girard [73] introduces a third- party architecture, communication overhead is increased that may degrade the netw ork performance. So, Kim et al. [74] propose a dual ke y based activ ation scheme to support key generation and update without adding any complexities. This paper explicates that AppK ey which is not updated periodically will pose many security threats through which an attacker can steal all transmitted data of a target node. The proposed technique of this paper separates the Network and Application Session key generation to appropriate servers. These keys cannot be deri ved from a public key and not shared with other devices. This scheme is proven to be both delay and po wer feasible through e xperiments. T rust mec hanism and blockc hains. Some works [131]– [136], discuss the application of blockchain technology for 15 IoT scenarios. This is helpful to apply blockchain to LoRa networks. Lin et al. [137] build a trust mechanism for LoRa using blockchain technology as attacking a blockchain system is computationally difﬁcult as the attacker has to transcend at least half of the system’ s computational ability . The pro- posed system implements Blockchain manager component to network server . This frame work is proposed for large scale deployments of LoRa like wild-life monitoring, asset tracking and smart parking. Key Insights. The key insights regarding the security issues are listed as follows. • An attacker can decrypt messages in the network by compromising security keys if they can get physical access to end-de vices. • LoRa de vices are susceptible to jamming attack, replay attack, beacon synchronization attack, traf ﬁc analysis and man-in-the-middle attack. • Larger size DevNonce prev ents join attacks [76]. • Application and Network key generation and update has to be separated based on the application scenario. • Blockchain mechanisms can also be implemented on Low powered devices [137]. V I . O P E N I S S U E S As introduced in V, v arious techniques have been proposed to address the challenges of LoRa deployments. Some so- lutions still leave room to further improve the performance of LoRa. For example, some solutions for choosing dynamic communication parameters consider most of the factors, but did not take into account the ambient temperature which plays a major role in reducing the signal strength [68]. Based on the abov e analysis of research challenges and recently proposed solutions in section IV and V, some open issues of the LoRa technology are presented in this section. A. Optimal placement of multiple gateways: Some works, [38], [21], [66], [69] aiding scalability and interference use multiple gatew ays as a solution. Even though these techniques outperform existing results, using multiple gate ways instigate to study the optimal gateway placements for LoRa deployments. Optimal placement of gate ways is always dependent on the application and constraints of the hardware used in the application. A generic solution for optimal placement of gate ways according to the categories of applications will further improve the performance. B. Link Co-or dination: Countering degradation due to downlink A CKs: Some works like [39] and [64] state that do wnlink ACKs reduce network performance, as end devices are not able to transmit subsequent packets if do wnlink A CKs are delayed or cor- rupted. This giv es rise to the need for dev eloping a dynamic A CK mechanism to improv e network performance. Dynamic retransmission policies: One solution to counter downlink A CKs will be setting dynamic retransmission poli- cies. Static retransmission policy degrades the network perfor- mance when the time taken by an A CK to reach the end node is larger than the retransmission time. This explains the need for dynamic retransmission policies. A joint retransmission policy with channel coding and compression is theoretically studied, but the implementation of such computationally expensiv e techniques on constrained LoRa hardware must be considered in future. Besides, it has to be noted that a modular retransmis- sion policy without any dependencies on other techniques is inevitable. None of the techniques has addressed this problem of v arying retransmission timers dynamically . The factors triggering retransmission, ev en in the case of correct reception, must be extensiv ely studied through experiments for various scenarios and a dynamic policy must be devised to improv e the performance of LoRa. C. Communication Range Performance ev aluations in [32]–[37] expound the need to provide reliable transmissions for long range LoRa links. Some of the techniques lik e Du et al. [96] improves communication range for wireless sensor networks without hea vy hardware modiﬁcations. The communication range could be further enlarged in the future as LoRa chips will ev olve to support new functionalities. Choir [51] improv es the range but its implementation on commercial radio chips is ambiguous as they were experi- mented on USRP radios and may require modifying the com- mercial radio. NetScatter [55] theoretically proves that Choir [51] cannot scale-up well. NetScatter [55] le verages backscat- ter with distributed coding to enable concurrent transmission of 256 nodes which still needs some hardware modiﬁcations on commercial chips. Ev en though the abov e techniques Choir [51] and NetScatter [55] hav e improved the range, communication range still needs to be improv ed without heavy modiﬁcations of the commercially available chips . D. Security Security spans ov er a range of attacks like node imper- sonation, eavesdropping, Black hole attack, W ormhole attack, etc., as discussed by Zhou et al. [86]. Only few key man- agement techniques have been discussed and proposed for LoRa. Several attacks still need to be addressed to secure LoRa networks. Even though some techniques enhance the security of e xisting LoRa standard, security requirements of LoRa are not discussed based on the applications of LoRa. In the future, each application of LoRa will demand their own security needs. For example, some applications may require Network and Application Session keys to be independent and Network Session K ey should be conﬁdential from application server and vice-versa. Hence, the security needs of each deploying scenario have to be deeply in vestigated to mitigate vulnerabilities arising due to different application scenarios. V I I . C O N C L U S I O N Among dif ferent LPW AN technologies, LoRa networking is widely adopted, since it allows to build and maintain an autonomous network without third-party infrastructure, while satisfying the low power and long range communication 16 requirements. By in vestigating the challenges faced during deploying LoRa networks, recent solutions developed are dis- cussed in detail. Based on the challenges and recent solutions, we present some open issues that need to be addressed for practical lar ge-scale deployment of LoRa networks. R E F E R E N C E S [1] J. Lin, W . Y u, N. Zhang, X. Y ang, H. Zhang, and W . Zhao, “ A survey on internet of things: Architecture, enabling technologies, security and priv acy , and applications, ” IEEE Internet of Things Journal , vol. 4, no. 5, pp. 1125–1142, 2017. [2] A. Al-Fuqaha, M. Guizani, M. Mohammadi, M. Aledhari, and M. A yyash, “Internet of things: A survey on enabling technologies, protocols, and applications, ” IEEE Communications Surve ys & T utori- als , vol. 17, no. 4, pp. 2347–2376, 2015. [3] J. A. Stankovic, “Research directions for the internet of things, ” IEEE Internet of Things Journal , vol. 1, no. 1, pp. 3–9, 2014. [4] “5G americas L TE progress leading to the 5G massive Internet of Things, ” 2017. [5] “Northstream white paper, Massiv e IoT : Different technologies for different needs, ” 2017. [6] Ericsson, “Ericsson mobility report, ” 2017. [7] V . Gazis, “ A surve y of standards for machine-to-machine and the internet of things, ” IEEE Communications Surveys & T utorials , vol. 19, no. 1, pp. 482–511, 2017. [8] S. Chen, R. Ma, H.-H. Chen, H. Zhang, W . Meng, and J. Liu, “Machine-to-machine communications in ultra-dense networksa sur- vey , ” IEEE Communications Surveys & T utorials , vol. 19, no. 3, pp. 1478–1503, 2017. [9] U. Raza, P . Kulkarni, and M. Sooriyabandara, “Low power wide area networks: An overvie w , ” IEEE Communications Surveys & T utorials , vol. 19, no. 2, pp. 855–873, 2017. [10] H. W ang and A. O. Fapojuwo, “ A surve y of enabling technologies of low power and long range machine-to-machine communications, ” IEEE Communications Surve ys & T utorials , vol. 19, no. 4, pp. 2621–2639, 2017. [11] D. Flore, “3GPP standards for the Internet-of-Things, ” Recuperado el , vol. 25, 2016. [12] J. C. Zuniga and B. Ponsard, “Sigfox system description, ” LPW AN@ IETF97, Nov . 14th , 2016. [13] L. V angelista, A. Zanella, and M. Zorzi, “Long-range IoT technologies: The dawn of LoRa, ” in Proceedings of the Futur e Access Enablers of Ubiquitous and Intelligent Infrastructures , 2015, pp. 51–58. [14] J. Lee, Y . Kim, Y . Kwak, J. Zhang, A. Papasakellariou, T . Novlan, C. Sun, and Y . Li, “L TE-advanced in 3GPP rel-13/14: An ev olution tow ard 5G, ” IEEE Communications Magazine , vol. 54, no. 3, pp. 36– 42, 2016. [15] M. Mezza villa, M. Zhang, M. Polese, R. Ford, S. Dutta, S. Rangan, and M. Zorzi, “End-to-end simulation of 5G mmwave networks, ” IEEE Communications Surve ys & T utorials , vol. 20, no. 3, pp. 2237–2263, 2018. [16] R. S. Sinha, Y . W ei, and S.-H. Hwang, “ A survey on LPW A technology: LoRa and NB-IoT, ” ICT Expr ess , vol. 3, no. 1, pp. 14–21, 2017. [17] Ingenu, “How RPMA works?” White P aper , 2016. [18] W . W ebb, Understanding W eightless: T echnology , Equipment, and Network Deployment for M2M Communications in White Space . Cam- bridge University Press, 2012. [19] LoRa-Alliance, “ A technical overvie w of LoRaW AN, ” T echnical Mar- keting W orkgroup, Nov . 2015. [20] R. Fujdiak, P . Blazek, K. Mikhaylov , L. Malina, P . Mlynek, J. Misurec, and V . Blazek, “On track of sigfox conﬁdentiality with end-to-end encryption, ” in Pr oceedings of the 13th ACM International Conference on A vailability , Reliability and Security , 2018, p. 19. [21] M. C. Bor , U. Roedig, T . V oigt, and J. M. Alonso, “Do LoRa lo w- power wide-area networks scale?” in Proceedings of the International Confer ence on Modeling , Analysis and Simulation of Wir eless and Mobile Systems , 2016, pp. 59–67. [22] Semtech, “Smart buildings transformed using semtech’ s LoRa technol- ogy , ” Apr. 2017, white paper . [23] Semtech, “Smart cities transformed using semtech’ s LoRa technology , ” Jul. 2017, white paper. [24] Semtech, “Rev olutionising smart agriculture using semtech’ s LoRa technology , ” Oct. 2017, white paper . [25] Semtech, “Real-world LoRaW AN network capacity for electrical me- tering applications, ” Sep. 2017, white paper . [26] D. Magrin, M. Centenaro, and L. V angelista, “Performance evaluation of LoRa networks in a smart city scenario, ” in Proceedings of the International Conference On Communications , 2017, pp. 1–7. [27] W . Du, Z. Xing, M. Li, B. He, L. H. C. Chua, and H. Miao, “Optimal sensor placement and measurement of wind for water quality studies in urban reservoirs, ” in Pr oceedings of the International symposium on Information pr ocessing in sensor networks , 2014, pp. 167–178. [28] W . Du, Z. Xing, M. Li, B. He, L. H. C. Chua, and H. Miao, “Sensor placement and measurement of wind for water quality studies in urban reservoirs, ” A CM Tr ansactions on Sensor Networks , vol. 11, no. 3, pp. 41:1–41:27, 2015. [29] W . Du, J. C. Liando, H. Zhang, and M. Li, “Pando: Fountain-enabled fast data dissemination with constructiv e interference, ” IEEE/ACM T ransactions on Networking , vol. 25, no. 2, pp. 820–833, 2017. [30] W . Du, J. Liando, H. Zhang, and M. Li, “When pipelines meet fountain: Fast data dissemination in wireless sensor networks, ” in Proceedings of the ACM Conference on Embedded Networked Sensor Systems , 2015, pp. 365–378. [31] J.Catalano, J-P .Coupigny , M.Kuyper , N.sornin, and A.Y egin, “Fuota process summary technical recommendation, ” 2019. [32] R. Fuidiak, P . Mlynek, J. Misurec, and M. Strajt, “Simulated cov erage estimation of single gateway LoRaW AN network, ” in Proceedings of the International Conference on Systems, Signals and Image Process- ing , 2018, pp. 1–4. [33] J. Petajajarvi, K. Mikhaylov , A. Roivainen, T . Hanninen, and M. Pet- tissalo, “On the co verage of LPW ANs: range e valuation and channel attenuation model for LoRa technology , ” in Pr oceedings of the Inter- national Conference on ITS T elecommunications , 2015, pp. 55–59. [34] J. Pet ¨ aj ¨ aj ¨ arvi, K. Mikhaylov , R. Y asmin, M. H ¨ am ¨ al ¨ ainen, and J. Iinatti, “Evaluation of LoRa LPW AN technology for indoor remote health and wellbeing monitoring, ” International Journal of W ireless Information Networks , vol. 24, no. 2, pp. 153–165, 2017. [35] N. Harris and J. Curry , “Development and range testing of a LoRaW AN system in an urban environment, ” W orld Academy of Science, Engi- neering and T echnology , International Journal of Electrical, Computer , Ener getic, Electronic and Communication Engineering , vol. 12, no. 1, pp. 43–51, 2018. [36] J. Haxhibeqiri, A. Karaa ˘ gac ¸ , F . V an den Abeele, W . Joseph, I. Moer- man, and J. Hoebeke, “LoRa indoor coverage and performance in an industrial en vironment: Case study , ” in Pr oceedings of the International Confer ence on Emerging T echnologies And F actory Automation , 2017, pp. 1–8. [37] K. Mikhaylov , J. Petaejaejaervi, and T . Haenninen, “ Analysis of capacity and scalability of the LoRa low power wide area network technology , ” in Pr oceedings of the Eur opean W ireless Confer ence , 2016, pp. 1–6. [38] Semtech, “LoRaW AN’ s capacity trials in dense urban environments. ” [39] D. Bankov , E. Khorov , and A. L yakhov , “On the limits of LoRaW AN channel access, ” in Pr oceedings of the International Conference on Engineering and T elecommunication , 2016, pp. 10–14. [40] L. Angrisani, P . Arpaia, F . Bona volont ` a, M. Conti, and A. Liccardo, “LoRa protocol performance assessment in critical noise conditions, ” in Pr oceedings of the International F orum on Research and T echnologies for Society and Industry , 2017, pp. 1–5. [41] N. Blenn and F . Kuipers, “LoRaW AN in the wild: Measurements from the things network, ” arXiv pr eprint arXiv:1706.03086 , 2017. [42] M. Lauridsen, B. V ejlgaard, I. Z. Ko vacs, H. Nguyen, and P . Mogensen, “Interference measurements in the European 868 MHz ISM band with focus on LoRa and SigFox, ” in Pr oceedings of the W ir eless Communications and Networking Conference , 2017, pp. 1–6. [43] B. V ejlgaard, M. Lauridsen, H. Nguyen, I. Z. K ov ´ acs, P . Mogensen, and M. Sorensen, “Coverage and capacity analysis of Sigfox, LoRa, GPRS, and NB-IoT, ” in Proceedings of the V ehicular T echnology Conference , 2017, pp. 4–7. [44] G. Zhu, C.-H. Liao, M. Suzuki, Y . Narusue, and H. Morikawa, “Evaluation of LoRa recei ver performance under co-technology inter- ference, ” in Pr oceedings of the Annual Consumer Communications & Networking Conference , 2018, pp. 1–7. [45] C. Orfanidis, L. M. Feeney , M. Jacobsson, and P . Gunningberg, “Inv es- tigating interference between LoRa and IEEE 802.15.4g networks, ” in Pr oceedings of the W ir eless and Mobile Computing, Networking and Communications , 2017, pp. 1–8. [46] G. Ferr ´ e, “Collision and packet loss analysis in a LoRaW AN network, ” in Pr oceedings of the Eur opean Signal Pr ocessing Conference , 2017, pp. 2586–2590. 17 [47] E. Aras, G. S. Ramachandran, P . Lawrence, and D. Hughes, “Exploring the security vulnerabilities of LoRa, ” in Pr oceedings of the Interna- tional Conference on Cybernetics , 2017, pp. 1–6. [48] I. Butun, N. Pereira, and M. Gidlund, “ Analysis of LoRaW AN v1. 1 security , ” in Proceedings of the MobiHoc W orkshop on Experiences with the Design and Implementation of Smart Objects , 2018, pp. 5–10. [49] R. Miller, “Lora security: Building a secure lora solution, ” MWR Labs Whitepaper , 2016. [50] B. Oniga, V . Dadarlat, E. De Poorter , and A. Munteanu, “ Analysis, design and implementation of secure LoRaW AN sensor networks, ” in Pr oceedings of the International Confer ence on Intelligent Computer Communication and Processing , 2017, pp. 421–428. [51] R. Eletreby , D. Zhang, S. Kumar , and O. Y a ˘ gan, “Empowering low- power wide area networks in urban settings, ” in Proceedings of the Confer ence of the Special Inter est Group on Data Communication , 2017, pp. 309–321. [52] A. Dongare, R. Narayanan, A. Gadre, A. Luong, A. Balanuta, S. Ku- mar , B. Iannucci, and A. Ro we, “Charm: Exploiting geographical div er- sity through coherent combining in low-po wer wide-area networks, ” in Pr oceedings of the International Confer ence on Information Processing in Sensor Networks , 2018, pp. 60–71. [53] Y . Peng, L. Shangguan, Y . Hu, Y . Qian, X. Lin, X. Chen, D. F ang, and K. Jamieson, “PLoRa: a passive long-range data network from ambient lora transmissions, ” in Pr oceedings of the Conference of the Special Interest Group on Data Communication , 2018, pp. 147–160. [54] V . T alla, M. Hessar , B. K ellogg, A. Najaﬁ, J. R. Smith, and S. Gol- lakota, “LoRa backscatter: Enabling the vision of ubiquitous con- nectivity , ” Proceedings of the on Interactive, Mobile, W earable and Ubiquitous T echnologies , vol. 1, no. 3, p. 105, 2017. [55] M. Hessar , A. Najaﬁ, and S. Gollakota, “Netscatter: Enabling large- scale backscatter networks, ” arXiv pr eprint arXiv:1808.05195 , 2018. [56] P . J. Marcelis, V . S. Rao, and R. V . Prasad, “DaRe: Data recovery through application layer coding for LoRaW AN, ” in Proceedings of the International Confer ence on Internet-of-Things Design and Imple- mentation , 2017, pp. 97–108. [57] O. Geor giou and U. Raza, “Low po wer wide area network analysis: Can LoRa scale?” IEEE W ireless Communications Letters , v ol. 6, no. 2, pp. 162–165, 2017. [58] T . C. D ¨ onmez and E. Nigussie, “Security of LoRaW AN v1. 1 in backward compatibility scenarios, ” Pr ocedia Computer Science , vol. 134, pp. 51–58, 2018. [59] M. Sandell and U. Raza, “ Application layer coding for iot: beneﬁts, lim- itations, and implementation aspects, ” IEEE Systems Journal , no. 99, pp. 1–8, 2018. [60] J. Haxhibeqiri, F . V an den Abeele, I. Moerman, and J. Hoebeke, “LoRa scalability: A simulation model based on interference measurements, ” Sensors , vol. 17, no. 6, p. 1193, 2017. [61] B. Reynders, Q. W ang, P . T uset-Peiro, X. V ilajosana, and S. Pollin, “Improving reliability and scalability of LoRaW ANs through lightweight scheduling, ” IEEE Internet of Things Journal , 2018. [62] K. Q. Abdelfadeel, V . Cionca, and D. Pesch, “ A fair adaptiv e data rate algorithm for LoRaW AN, ” arXiv pr eprint arXiv:1801.00522 , 2018. [63] B. Reynders, W . Meert, and S. Pollin, “Power and spreading factor control in low po wer wide area networks, ” in Proceedings of the International Conference on Communications , 2017, pp. 1–6. [64] A.-I. Pop, U. Raza, P . Kulkarni, and M. Sooriyabandara, “Does bidirectional traf ﬁc do more harm than good in LoRaW AN based LPW A networks?” arXiv preprint , 2017. [65] F . Cuomo, M. Campo, A. Caponi, G. Bianchi, G. Rossini, and P . Pisani, “EXPLoRa: Extending the performance of LoRa by suitable spreading factor allocations, ” in Pr oceedings of the W ireless and Mobile Computing, Networking and Communications , 2017, pp. 1–8. [66] F . V an den Abeele, J. Haxhibeqiri, I. Moerman, and J. Hoebeke, “Scalability analysis of large-scale LoRaW AN networks in NS-3, ” Internet of Things Journal , vol. 4, no. 6, pp. 2186–2198, 2017. [67] K. Mikhaylov , J. Pet ¨ aj ¨ aj ¨ arvi, and J. Janhunen, “On LoRaW AN scala- bility: Empirical evaluation of susceptibility to inter -network interfer- ence, ” in Pr oceedings of the Eur opean Confer ence on Networks and Communications , 2017, pp. 1–6. [68] M. Cattani, C. A. Boano, and K. R ¨ omer , “ An experimental e valuation of the reliability of LoRa long-range low-power wireless communication, ” Journal of Sensor and Actuator Networks , vol. 6, no. 2, p. 7, 2017. [69] T . V oigt, M. Bor , and U. Roedig, “Mitigating inter-network interference in LoRa networks, ” in Proceedings of the MadCom: New W ireless Communication P aradigms for the Internet of Things , 2017, pp. 323– 328. [70] J. Lee, D. Hwang, J. Park, and K.-H. Kim, “Risk analysis and countermeasure for bit-ﬂipping attack in LoRaW AN, ” in Pr oceedings of the International Conference on Information Networking , 2017, pp. 549–551. [71] J. Kim and J. Song, “ A simple and efﬁcient replay attack prevention scheme for LoRaW AN, ” in Pr oceedings of the International Confer ence on Communication and Network Security , 2017, pp. 32–36. [72] S. Na, D. Hwang, W . Shin, and K.-H. Kim, “Scenario and countermea- sure for replay attack using join request messages in LoRaW AN, ” in Pr oceedings of the International Conference on Information Network- ing , 2017, pp. 718–720. [73] P . Girard, “Low power wide area network security , ” gematlo, T ech. Rep., Dec. 2015. [74] J. Kim and J. Song, “ A dual ke y-based activ ation scheme for secure LoRaW AN, ” W ireless Communications and Mobile Computing , vol. 2017, pp. 1–12, 2017. [75] S. Naoui, M. E. Elhdhili, and L. A. Saidane, “Enhancing the security of the IoT LoRaW AN architecture, ” in Proceedings of the International Confer ence on P erformance Evaluation and Modeling in Wir ed and W ireless Networks , 2016, pp. 1–7. [76] S. T omasin, S. Zulian, and L. V angelista, “Security analysis of Lo- RaW AN join procedure for internet of things networks, ” in Pr oceedings of the W ireless Communications and Networking Conference W ork- shops , 2017, pp. 1–6. [77] J. Haxhibeqiri, E. De Poorter, I. Moerman, and J. Hoebeke, “ A survey of lorawan for iot: From technology to application, ” Sensors , vol. 18, no. 11, p. 3995, 2018. [78] S. Demetri, M. Z ´ u ˜ niga, G. P . Picco, F . Kuipers, L. Bruzzone, and T . T elkamp, “ Automated estimation of link quality for lora: A remote sensing approach, ” in Pr oceedings of the International Conference on Information Pr ocessing in Sensor Networks (IPSN) , 2019. [79] R. El Chall, S. Lahoud, and M. El Helou, “Loraicdcsnetwork: Radio propagation models and performance ev aluation in various environ- ments in lebanon, ” IEEE Internet of Things Journal , 2019. [80] J. C. Liando, A. Gamage, A. W . T engourtius, and M. Li, “Known and unknown facts of lora: Experiences from a large-scale measurement study , ” A CM T ransactions on Sensor Networks (TOSN) , vol. 15, no. 2, p. 16, 2019. [81] Semtech, “Smart parking, ” https://www .semtech.com/lora/ lora- applications, White Paper . [82] Semtech, “Smart lighting, ” https://www .semtech.com/lora/ lora- applications, White Paper . [83] Semtech, “Lora technology: Transforming golf courses with iot, ” T ech. Rep., 2018. [84] Semtech, “Lora technology: Mallorca dev elops ﬁrst lorawan smart island, ” T ech. Rep., 2018. [85] Z. Qin and J. A. McCann, “Resource efﬁciency in Low-Power W ide- Area Networks for IoT applications, ” in Proceedings of the Global Communications Conference , 2017, pp. 1–7. [86] Y . Zhou, Y . Fang, and Y . Zhang, “Securing wireless sensor networks: a survey , ” IEEE Communications Surve ys & T utorials , vol. 10, no. 3, 2008. [87] F . Orfei, C. B. Mezzetti, and F . Cottone, “V ibrations po wered LoRa sensor: An electromechanical energy harvester working on a real bridge, ” in Proceedings of the Sensors , 2016, pp. 1–3. [88] A. V arshney , O. Harms, C. P ´ erez-Penichet, C. Rohner, F . Hermans, and T . V oigt, “Lorea: A backscatter architecture that achie ves a long communication range, ” in Pr oceedings of the ACM Conference on Embedded Network Sensor Systems , 2017, p. 18. [89] J. Navarro-Ortiz, S. Sendra, P . Ameigeiras, and J. M. Lopez-Soler, “Integration of LoRaW AN and 4G/5G for the Industrial Internet of Things, ” IEEE Communications Magazine , vol. 56, no. 2, pp. 60–67, 2018. [90] Ideetron, “LORANK gatew ay , ideetron lora products, ” url = https://webshop.ideetron.nl/LoRa. [91] W . Gao and W . Du and Z. Zhao and G. Min and M. Singhal, “T owards Energy-F airness in LoRa Networks, ” in IEEE ICDCS , 2019. [92] P . Hu, P . Zhang, M. Rostami, and D. Ganesan, “Braidio: An integrated activ e-passiv e radio for mobile de vices with asymmetric ener gy bud- gets, ” in Proceedings of the 2016 ACM SIGCOMM Confer ence . ACM, 2016, pp. 384–397. [93] WIMOD, “WIMOD - IM880A Data sheet, ” WiMOD iM880A , T ech. Rep., Jul. 2017. [94] Kerlink, “W irnet station datasheet, ” T ech. Rep., Jan. 2016. [95] Semtech, “Sx1272/3/6/7/8: Lora modem designers guide, ” Jul. 2013, revision 1. 18 [96] W . Du, Z. Li, J. C. Liando, and M. Li, “From rateless to distanceless: Enabling sparse sensor network deployment in large areas, ” IEEE/ACM T ransactions on Networking , vol. 24, no. 4, pp. 2498–2511, 2016. [97] W . Du, Z. Li, J. Liando, and M. Li, “From rateless to distanceless: enabling sparse sensor network deployment in large areas, ” in Pr o- ceedings of the 12th ACM Confer ence on Embedded Network Sensor Systems , 2014, pp. 134–147. [98] A. Saifullah, M. Rahman, D. Ismail, C. Lu, R. Chandra, and J. Liu, “Snow: Sensor network over white spaces, ” in A CM IPSN , 2016. [99] M. Rahman and A. Saifullah, “Integrating low-po wer wide-area net- works in white spaces, ” in Pr oceedings of the Third International Confer ence on Internet-of-Things Design and Implementation , 2018, pp. 255–260. [100] X. Xia, Y . Zheng, and T . Gu, “Ftrack: Parallel decoding for lora transmissions, ” in Pr oceedings of the A CM Confer ence on Embedded Networked Sensor Systems , 2019. [101] R. Huang, H. Li, B. Hamzeh, Y . Choi, S. Mohanty , and C. Hsu, “Proposal for evaluation methodology for 802.16 p. ” [102] N. Kouvelas, V . Rao, and R. Prasad, “Employing p-csma on a lora network simulator , ” arXiv pr eprint arXiv:1805.12263 , 2018. [103] M. Levorato, U. Mitra, and M. Zorzi, “Cognitive interference manage- ment in retransmission-based wireless networks, ” IEEE T ransactions on Information Theory , vol. 58, no. 5, pp. 3023–3046, 2012. [104] W .-H. K uo, R. Kaliski, and H.-Y . W ei, “ A qoe-based link adaptation scheme for h. 264/svc video multicast over ieee 802.11, ” IEEE T rans- actions on Cir cuits and Systems for V ideo T echnology , vol. 25, no. 5, pp. 812–826, 2015. [105] D. Chang, K. Cho, N. Choi, Y . Choi et al. , “ A probabilistic and op- portunistic ﬂooding algorithm in wireless sensor networks, ” Computer Communications , vol. 35, no. 4, pp. 500–506, 2012. [106] L. Xiong, L. Libman, and G. Mao, “Optimal strategies for cooperative mac-layer retransmission in wireless networks, ” in IEEE W ireless Communications and Networking Conference , 2008. [107] G. Naddafzadeh-Shirazi, P .-Y . Kong, and C.-K. Tham, “Distributed reinforcement learning frameworks for cooperative retransmission in wireless networks, ” IEEE Tr ansactions on V ehicular T echnology , vol. 59, no. 8, pp. 4157–4162, 2010. [108] C.-Y . W an, A. T . Campbell, and L. Krishnamurthy , “Psfq: a reliable transport protocol for wireless sensor networks, ” in Proceedings of the ACM international workshop on W ireless sensor networks and applications , 2002. [109] H. W en, C. Lin, F . Ren, Y . Y ue, and X. Huang, “Retransmission or redundancy: Transmission reliability in wireless sensor networks, ” in IEEE International Conference on Mobile Adhoc and Sensor Systems , 2007. [110] C. Pielli, C. Stefano vic, P . Popovski, and M. Zorzi, “Joint retransmis- sion, compression and channel coding for data ﬁdelity under energy constraints, ” arXiv preprint , 2017. [111] Semtech, “LoRaMote user guide, ” Jul. 2014. [112] O. Iova, A. Murphy , G. P . Picco, L. Ghiro, D. Molteni, F . Ossi, and F . Cagnacci, “LoRa from the city to the mountains: Exploration of hard- ware and environmental factors, ” in Pr oceedings of the International Confer ence on Embedded W ireless Systems and Networks , 2017. [113] C. Hakkenberg, “Experimental e valuation of LoRa(W AN) in indoor and outdoor environments, ” Master’ s thesis, University of T wente, 2016. [114] P . Neumann, J. Montavont, and T . No ¨ el, “Indoor deployment of lo w- power wide area networks (LPW AN): A LoRaW AN case study , ” in Pr oceesdings of the International Conference on W ireless and Mobile Computing, Networking and Communications , 2016, pp. 1–8. [115] A. Augustin, J. Yi, T . Clausen, and W . M. T ownsley , “ A study of LoRa: Long range & low power networks for the Internet of Things, ” Sensors , vol. 16, no. 9, p. 1466, 2016. [116] N. Semiconductors, “Frdm-17510evb evaluation board user guide, ” Apr . 2017, re vision 1. [117] Cisco, “Cisco 910 industrial router data sheet, ” Aug. 2017. [118] N. Sornin, M. Luis, T . Eirich, T . Kramp, and O. Hersent, “LoRaW AN speciﬁcation, ” LoRa alliance , 2015. [119] M. Hata, “Empirical formula for propagation loss in land mobile radio services, ” IEEE transactions on V ehicular T echnology , vol. 29, no. 3, pp. 317–325, 1980. [120] Y . Okumura, “Field strength and its v ariability in vhf and uhf land- mobile radio service, ” Rev . Electr . Commun. Lab. , v ol. 16, pp. 825–873, 1968. [121] P . E. Mogensen, P . Eggers, C. Jensen, and J. B. Andersen, “Urban area radio propagation measurements at 955 and 1845 mhz for small and micro cells, ” in IEEE Global T elecommunications Confer ence GLOBECOM’91: Countdown to the New Millennium. Conference Recor d . IEEE, 1991, pp. 1297–1302. [122] B. Scholkopf and A. J. Smola, Learning with kernels: support vector machines, re gularization, optimization, and beyond . MIT press, 2001. [123] M. Bor and U. Roedig, “LoRa transmission parameter selection, ” in Pr oceedings of the International Confer ence on Distributed Computing in Sensor Systems , 2017, pp. 5–7. [124] B. Reynders, W . Meert, and S. Pollin, “Range and coexistence analysis of long range unlicensed communication, ” in Proceedings of the International Conference on T elecommunications , 2016, pp. 1–6. [125] C. J. F . Cremers, Scyther: Semantics and veriﬁcation of security pr o- tocols . Eindhoven University of T echnology Eindhoven, Netherlands, 2006. [126] Y . B. Saied and A. Oliv ereau, “HIP tiny exchange (tex): A distributed key exchange scheme for HIP-based Internet of Things, ” in Pr oceed- ings of the Third International Conference on Communications and Networking , 2012, pp. 1–8. [127] Y . B. Saied, A. Olivereau, D. Zeghlache, and M. Laurent, “Lightweight collaborativ e ke y establishment scheme for the aInternet of Things, ” Computer Networks , vol. 64, pp. 273–295, 2014. [128] P . Porambage, A. Braeken, P . Kumar , A. Gurtov , and M. Ylianttila, “Proxy-based end-to-end ke y establishment protocol for the Internet- of-Things, ” in Pr oceedings of the International Confer ence on Com- munication W orkshop , 2015, pp. 2677–2682. [129] S.-H. Seo, J. W on, S. Sultana, and E. Bertino, “Effecti ve ke y man- agement in dynamic wireless sensor networks, ” IEEE T ransactions on Information F orensics and Security , vol. 10, no. 2, pp. 371–383, 2015. [130] S. Agrawal, R. Roman, M. L. Das, A. Mathuria, and J. Lopez, “ A novel key update protocol in mobile sensor networks, ” in Proceedings of the International Conference on Information Systems Security , 2012, pp. 194–207. [131] K. Christidis and M. Devetsikiotis, “Blockchains and smart contracts for the Internet of Things, ” IEEE Access , vol. 4, pp. 2292–2303, 2016. [132] K. Biswas and V . Muthukkumarasamy , “Securing smart cities using blockchain technology , ” in Pr oceedings of the International Confer ence on High P erformance Computing and Communications; International Confer ence on Smart City; International Confer ence on Data Science and Systems , 2016, pp. 1392–1393. [133] A. Dorri, S. S. Kanhere, R. Jurdak, and P . Gaurav aram, “Blockchain for IoT security and priv acy: The case study of a smart home, ” in Pr oceedings of the International Conference on P ervasive Computing and Communications W orkshops , 2017, pp. 618–623. [134] S. Huh, S. Cho, and S. Kim, “Managing IoT devices using blockchain platform, ” in Proceedings of the International Conference on Advanced Communication T echnology , 2017, pp. 464–467. [135] M. Samaniego and R. Deters, “Using blockchain to push software- deﬁned IoT components onto edge hosts, ” in Pr oceedings of the Inter- national Conference on Big Data and Advanced Wir eless T echnologies , 2016, p. 58. [136] R. D. Mayra Samaniego, “Blockchain as a service for IoT, ” in Pr oceedings of the International Conference on Internet of Things and Gr een Computing and Communications and Cyber , Physical and Social Computing and Smart Data , 2016, pp. 433–436. [137] J. Lin, Z. Shen, and C. Miao, “Using blockchain technology to build trust in sharing LoRaW AN IoT, ” in Pr oceedings of the International Confer ence on Crowd Science and Engineering , 2017, pp. 38–43.

A Survey on LoRa Networking: Research Problems, Current Solutions and Open Issues

Original Paper

Comments & Academic Discussion

Leave a Comment

Original Paper

Related Papers

Comments & Academic Discussion

Leave a Comment