GHOST - Safe-Guarding Home IoT Environments with Personalised Real-Time Risk Control
We present the European research project GHOST, (Safe-guarding home IoT environments with personalised real-time risk control), which challenges the traditional cyber security solutions for the IoT by proposing a novel reference architecture that is embedded in an adequately adapted smart home network gateway, and designed to be vendor-independent. GHOST proposes to lead a paradigm shift in consumer cyber security by coupling usable security with transparency and behavioural engineering.
KeywordsSmart home Security IoT Gateway Risk assessment
According to , the average IoT device was attacked once every two minutes in 2016. Unfortunately, such botnets as Mirai are taking advantage of the fact that security is still not a priority for device manufacturers, leading to the lack of possibility of automatic firmware upgrades, exposing the devices to simple attacks such as account enumeration and open ports scanning up to unpatched vulnerabilities presence and their exploitation to gain full control.
In addition to forcing the integration of security aspects into IoT devices at the manufacturer level, it is evident that a monitoring solution is essential to protect the end-users. IoT devices are often completely closed, not standardised or openly developed. Hence, the user does not have a clear idea of the potential risks involved. On top of the purely technological and operational cyber security challenges, the end-user behaviour becomes a determinant factor, with the human typically portrayed as the weakest link in security. Indeed, consumers tend to exhibit low tolerance and fatigue in using sophisticated cyber security solutions or practices, while the cyber security industry often addresses usability as at trade-off on security rather than as a security enhancing component. Thus, combining or integrating usability and security requirements is a major research challenge, which recently has been brought forward [2, 3], while turning end-user behaviour in favour of cyber security remains a field with a promising exploitation potential .
This paper gives an overview of the European Union Horizon 2020 Research and Innovation project GHOST (https://www.ghost-iot.eu/). GHOST aims to increase the level and the effectiveness of automation of existing cyber security services and to enhance system self-defence while prioritising the opening up the cyber security ‘blackbox’ to consumers and building trust through advanced usable transparency tools derived from end-users’ mental models.
The rest of the paper is structured as follows. Section 2 discusses related work. Section 3 presents the GHOST system, whilst Sect. 4 presents the GHOST validation process. Finally, conclusions are summarised in Sect. 5.
2 Related Work
In traditional cyber security, Intrusion Detection Systems (IDS) are taking the main role in detecting any anomalous activity on the network. Best known solutions are Snort , Suricata  and Bro . While Snort and Suricata are based on pattern matching detection, Bro is relying on semantic matching of the network events. However, these solutions are designed for professional use and are not explicitly aimed at the IoT environment in terms of protocol analysis availability. Global scale architecture with distributed data storage and correlation for IDS was proposed in . While taking advantage of novel technologies and providing wide coverage of monitored data for expert users, this system is not adapted for smart home installation where regular citizens have to understand the usage of this tool. Graphical representation of attack and threats scenes was greatly advanced in . These works are targeting professional analysts with the deep technology knowledge though. Modelling uncertainties in the cyber threat arena was presented in , Grey theory application for threat prediction was analysed in  and a framework assessing the impact of cyber attacks was described in . Once again, all these advancements are focusing on the expert users, not regular citizens.
2.1 Advancements in IoT Cyber Security Monitoring
Similarly to traditional cyber security IoT ecosystem is vulnerable to the analogous issues as in web, sensor and mobile communications networks, with particular focus on privacy, authentication and access control network configuration, information storage and management, standardisation and data integrity. The most complete classification of the IoT attack vectors is described in , referring to the IoT ecosystem as a Web3 or Web of Things phenomenon, where four main categories are provided: Device, Application Service, Network, Web Interface and Data integrity. Developing a cyber security solution targeting to protect all of the identified vectors is a very challenging and crucial task.  is raising a necessity to apply the Negative selection and Danger Theory to traditional IDS, to cover ubiquitous nature of the IoT devices and target all attack vectors specified above. Such systems, however, encounter serious limitations in terms computational power and storage requirements. An overview of the Real-time IoT security solutions was provided in . The authors conclude that existing approaches can be divided into two major classes: hardware and software based security. Alternative to IDS approach is described in , where SIEM system for IoT environment is proposed.
2.2 Smart Home Cyber Security Frameworks
The authors of  analyse existing architectures of smart homes from the security perspective, concluding that gateway architectures are the most suitable to provide key technologies for cyber protection: auto-configuration and automatic update installation. An overview of existing tooling for the implementation of cyber protection in smart homes is also included in their work, however, all these tools are applicable only for newly designed devices to be included in a future smart homes. On contrary, the IDS framework , based on Anomaly Behaviour Analysis, approaches this problem for existing and hardly changeable smart home installations. Their focus is given to measuring the activities of installed sensor devices a smart house is equipped with, and detecting any anomalies in the quantity and quality of the collected measurements. The limitation of their work relies in the ability to apply their analysis only on the primitive IoT devices without direct internet access. Similarly to GHOST, traffic monitoring and inspection solution IoTGuard, based on Bro, is presented in . The main drawback of their framework is the requirement to forward all router’s traffic to IoT Controller and link each IoT device with the IoT Watchdog. On the contrary, GHOST provides all-in-one solution to be deployed in the existing smart home installations with key focus given to user’s experience and understanding of a cyber security solution.
The great interest of developing smart home cyber security solutions is also given by the commercial entities. Already a wide selection of the commercial products is available on the market: F-Secure SENSE , Luma smart Wi-Fi router , Dojo , CUJO , Bitdefender , Norton Core .
3 The GHOST System
The GHOST system is being realised by analysing existing technical infrastructure and existing software components corresponding to the aims of the project. Usability studies have been defined with the aim to establish mental models of the end users. This allows systematical and effective addressing of the human factor with the aim to facilitate end-users’ proper decision making in relation to security and privacy issues and adequate usage of the GHOST solution. It also allows the definition of a first set of end-user requirements, which in turn facilitate better specification of the development and integration of core technologies. Since human participants will be involved in the evaluation phase of the project and personal data will be collected, special emphasis is given on elaborating a data management plan for respecting privacy related issues according to national and EU legislation. It should be noted that the access to the collected data will be provided only to the members of the consortium for development and demonstration purposes.
3.1 Development Approach
To keep up with cyber security issues and threats GHOST not only follows guidance documents, best practices and standards (issued by international, European and national stakeholders) at all stages of design and development, but it also scans for emerging threats/issues. To this end, it makes use of security intelligence available within the consortium and outside (e.g. through mining insightful security blogs), as well as related information collected directly from the end-users and the smart home pilots. The development of GHOST follows an iterative approach. Three iterations have been specified for the implementation of the technical components of the infrastructure. These will be evaluated through real life trials and feedback will be reflected back for further refinement and acceptance, according to the validation process discussed in Sect. 4.
3.2 GHOST Software Architecture
GHOST’s conceptual design involves advanced data flow analysis on a packet basis to build the context of communication. From this context, data are classified into user and device profiles, which in turn are used in the automated real-time risk assessment. The assessment is based on evaluation, comparison and matching with safe data flow patterns, utilising a self-learning approach. Data analytics and visualisation techniques are deployed to ensure enhanced user awareness and understanding of the security status, potential threats, risks, associated impacts and mitigation guidelines.
3.3 Core Layers
Data Interception and Inspection (DII). Data related to traffic of all network interfaces in a smart home environment is gathered directly from the network. This data is analysed and stored in order to be used by GHOST components. Significant data extracted from traffic packets is stored to a shared data storage. Additionally traffic packets are aggregated into groups related to specific communications or actions. These groups of packets are also analysed to extract information of a higher abstraction level and store it along with the information produced by single packets analysis. Additionally context information is extracted from traffic data. Recurring patterns of traffic are detected and the causes they are produced by are identified and an initial classification of the data type of traffic is performed. The network traffic may be correlated to actions of people or events in the smart home and the data in the packets are categorised accordingly as personal data or device data.
Contextual Profiling (CP). The classification templates and actual profiles of the typical devices’ behaviour are built in this layer, by extracting valuable data from the local network communication already prepared by DII profiles for the normal behaviour of the devices are built in a tree based format for further processing by the risk assessment component. This layer also monitors the communications occurring between any combination of devices including the gateway, along with the status of each device and the status of the gateway. Monitoring is learning based, and models are trained to recognise the normal status of devices and the normal status of communication between them. Random Neural Networks are employed for each pair of devices and reinforcement learning is used to update them through time.
Risk Assessment (RA). This a core layer, which gathers information about the current risks and analyses in real-time current network traffic flows. It correlates device activity on the network with the profiles available from CP layer. The automatic decision making of the Risk Engine presents transparency of the cyber security solution, informing the end-user only about urgent decisions. Its capabilities is enhanced with the use of Smart Contracts (SC) to ensure the reliability and trustworthiness of decisions. RA is also designed for controlling users’ privacy and making them aware of the associated risks.
Control and Monitoring (CM). Three types of the user interfaces are forming this layer: Feedback Analytics (advanced professional-alike interfaces), Security Intervention (daily decision-making support tooling) and Configuration. The input data include historical and current packet flow behaviours, risk levels, device profiles, packet classification score, etc. The layer provides visual and intuitive presentations and reports of the smart home security status, including visualisations of packet features through time, visual monitoring and distinction of packet behaviours, and visual identification of potential anomalies and vulnerabilities. The appropriate visualisation and human-machine interaction mechanisms are put in place to allow users to effortlessly and effectively review security issues and take key decisions that affect their privacy and security.
3.4 Supplementary Layers
Blockchain Defense Infrastructure (BDI). GHOST uses blockchain technology and SC for ensuring data and code integrity. At this layer the decisions made by RA are verified according to commonly agreed SC, turning the decision making into a truly decentralised and resilient system against intrusions. The integrity of the code running on smart home gateways can be certified by the use of blockchain technology. Additionally valuable security related information can be stored at a blockchain infrastructure in order to be shared between smart home gateways.
Cross Layer Anomaly Detection Framework (CLADF). Cross layer anomaly detection framework integrates existing open source solutions for traditional cyber security features. The main purpose is to collect, correlate, combine, and provide a unified output to other components in terms of possible events that require further analysis.
Cyber Security Knowledge Base (CSKB). A common cloud based knowledge repository is integrated with GHOST to collect anonymised security intelligence and insights from external web-sources to enhance the automatic decision making and improve end-user visual experience within the CM layer. It will maintain list of malicious actors and properties (IP Addresses, Domains, URLs, File Hashes).
Shared Data Storage (SDS). The data structures defined by each of the components are normalised and unified within a single storage framework. A combination of relational and non-relational databases is used to satisfy the needs of all components. There is distinction between local and cloud based storage, as some components will perform off-site analytics.
3.5 GHOST Hardware Platforms
maximum code complexity and size,
size of the memory of the system,
processing power that the device can offer in a certain period of time,
allowed energy consumption or battery duration,
communication methods and interfaces of the system,
user interfaces and accessibility to the system in deployment.
Several techniques has been proposed in the literature to keep these set of constraints controlled in different environments and specific solutions [27, 28, 29], including the for security and privacy applications [30, 31].
GHOST is being developed and tested using two resource-constrained platforms: a proprietary IoT gateway, and a Raspberry Pi (with some expansion modules for IoT networks). The use of both devices allows several different IoT protocols, such as 802.11, Bluetooth Low Energy, Z-Wave and 802.15.4 to run on GHOST. Differences do exist between these two devices, but there are also some similarities regarding their constraints as regards processing power; memory; communications; and energy efficiency. These constraints pose a number of research challenges.
4 GHOST Validation Process
The validation strategy defined for GHOST is based on a three-fold vision that combines a complete set of robustness and laboratory testing; the specific definition of realistic testbeds; and real-life trials or pilots. First, the laboratory testing will be done with the objectives of reducing the number of possible bugs and functional errors and of checking the stability of the hardware. Therefore, unit tests will be performed over each specific GHOST module and an acceptance test plan will be defined and tested, including both software and hardware stability testing. After this first stage, two already functional testbeds will be used to deeply test the functionality of the GHOST solution in a controlled environment. The testbeds designed for two specific smart home demonstrators include more than 15 different types of devices, involving up to 25 devices that will be simultaneously connected and monitored by the GHOST suite. In order to have a broad view of the possible services and solutions, devices like smart locks, biomedical devices, companion robots or smart lights based on several communication solutions (like 802.11, 802.15.4, Z-Wave or Bluetooth Low Energy) have been included in the testbeds. Potential threats against the smart home can be categorised into : (i) Physical attacks, (ii) Unintentional damage (accidental), (iii) Disaster (natural/environmental), (iv) Damages or loss of IT assets, (v) Failures/malfunctions, (vi) Outages, (vii) Eavesdropping/interception/hijacking, (viii) Nefarious activity/abuse, and (ix) Legal. Of these, relevant to GHOST are groups (ii), (iv), (vii), and (viii). Each of these groups includes a number of threats that can exploit relevant vulnerabilities by launching different attacks. The response of GHOST when faced with those amongst the above attacks that lead to higher risks and/or are most prevalent will be assessed in the controlled environment of the GHOST testbeds.
In addition to the testbeds, a set of pilots in real scenarios (homes of end-users) in three different countries (Spain, Romania and Norway) and with complementary use cases related with telecare, eHealth, home security and home automation will be carried out. The real-life trials have been designed to cover a varied set of application and services. Four different use cases have been defined: Ambient assisted living in smart homes for older people in Galicia, Spain; Continuous health monitoring for adult people in Galicia, Spain; Regular private homes (smart-home solutions) in Norway and Regular private homes (smart-home solutions) in Romania.
Each use case has their own set of devices to be installed and a complete test plan is being developed to simulate the possible results of specific attacks (previously validated and performed in the testbeds) to capture the response of the users to the GHOST behaviour.
GHOST brings professional security tools down to regular home users. The strategic outcome of GHOST is threefold: increased resilience of existing cyber security solutions for smart homes and the IoT; a leap forward to usability and automation in cyber security; and a boost in the competitiveness of European ICT security industry in the advent of the IoT in the connected world. From a user perspective, GHOST will help end-users to increase their control over their smart-home IoT devices and it will provide an option for smart-living service providers to use its security services to ensure that they respect the security and privacy needs of their clients.
Future work includes the iterative implementation, testing and validation of GHOST in existing laboratory testbeds and in real-life and scale pilots in three European countries, using appropriately designed use case scenarios. Related work from the GHOST project can be also found in [33, 34, 35].
This work is partially funded by the European Union’s Horizon 2020 Research and Innovation Programme through GHOST project (https://www.ghost-iot.eu/) under Grant Agreement No. 740923.
- 1.Chandrasekar, K., Cleary, G., Cox, O., Lau, H., Nahorney, B., Gorman, B.O., O’Brien, D., Wallace, S., Wood, P., Wueest, C.: ISTR April 2017. Internet Security Threat Report - Symantec 22(April), 77 (2017)Google Scholar
- 2.Nurse, J.R., Creese, S., Goldsmith, M., Lamberts, K.: Guidelines for usable cybersecurity: past and present. In: Proceedings of the 2011 3rd International Workshop on Cyberspace Safety and Security, CSS 2011, pp. 21–26 (2011)Google Scholar
- 3.Realpe, P.C., Collazos, C.A., Hurtado, J., Granollers, A.: Towards an integration of usability and security for user authentication. In: Proceedings of the XVI International Conference on Human Computer Interaction, p. 43. ACM (2015)Google Scholar
- 5.Roesch, M.: Snort: lightweight intrusion detection for networks. In: 13th Systems Administration Conference on LISA 1999, pp. 229–238 (1999)Google Scholar
- 6.(OISF), Open Information Security Foundation: Suricata. https://suricata-ids.org/
- 8.Marchal, S., Jiang, X., State, R., Engel, T.: A big data architecture for large scale security monitoring. In: Proceedings of the 2014 IEEE International Congress on Big Data, BigDataCongress 2014, pp. 56–63 (2014)Google Scholar
- 9.Koike, H., Ohno, K., Koizumi, K.: Visualizing cyber attacks using IP matrix. In: 2005 Proceedings of the IEEE Workshop on Visualization for Computer Security, VizSEC 2005, pp. 91–98 (2005)Google Scholar
- 10.Xie, P., Li, J.H., Ou, X., Liu, P., Levy, R.: Using Bayesian networks for cyber security analysis. In: 2010 IEEE/IFIP International Conference on Dependable Systems and Networks DSN, pp. 211–220 (2010)Google Scholar
- 11.Jibao, L., Huiqiang, W., Liang, Z.: Study of network security situation awareness model based on simple additive weight and grey theory. In: 2006 International Conference on Computational Intelligence and Security, vol. 2, pp. 1545–1548 (2006)Google Scholar
- 12.Jakobson, G.: Mission cyber security situation assessment using impact dependency graphs. In: 14th International Conference on Information Fusion, pp. 1–8, July 2011Google Scholar
- 14.Pamukov, M.E.: Application of artificial immune systems for the creation of IoT intrusion detection systems. In: 2017 9th IEEE International Conference on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications (IDAACS), pp. 564–568. IEEE, September 2017Google Scholar
- 15.Chen, C.Y., Hasan, M., Mohan, S.: Securing real-time internet-of-things. arXiv preprint arXiv:1705.08489, 1–10, May 2017
- 16.Zegzhda, P.: Safe integration of SIEM systems with internet of things: data aggregation, integrity control, and bioinspired safe routing. In: Proceedings of the 9th International Conference on Security of Information and Networks, SIN 2016, pp. 81–87 (2016)Google Scholar
- 18.Pacheco, J., Hariri, S.: IoT security framework for smart cyber infrastructures. In: 2016 IEEE 1st International Workshops on Foundations and Applications of Self* Systems (FAS*W), pp. 242–247. IEEE, September 2017Google Scholar
- 19.Park, Y., Daftari, S., Inamdar, P., Salavi, S., Savanand, A., Kim, Y.: IoTGuard: scalable and agile safeguards for internet of things. In: Proceedings of the IEEE Military Communications Conference MILCOM, pp. 61–66 (2016)Google Scholar
- 20.F-Secure: F-secure sense router. https://www.f-secure.com/en/web/home_global/sense
- 21.Luma Home Inc. https://lumahome.com/
- 22.BullGuard: Dojo by bullguard. https://dojo.bullguard.com/
- 23.CUJO: Cujo llc. https://www.getcujo.com/
- 24.Bitdefender BOX 2. https://www.bitdefender.com/box/
- 25.Norton Core™: Symantec corporation. https://us.norton.com/core
- 26.Bormann, C., Ersue, M., Keranen, A.: Terminology for constrained-node networks. Technical report, Internet Engineering Task Force (IETF), May 2014Google Scholar
- 30.Sethi, M., Kortoci, P., Di Francesco, M., Aura, T.: Secure and low-power authentication for resource-constrained devices. In: 2015 5th International Conference on the Internet of Things (IOT), pp. 30–36. IEEE, October 2015Google Scholar
- 31.Porambage, P., Braeken, A., Gurtov, A., Ylianttila, M., Spinsante, S.: Secure end-to-end communication for constrained devices in IoT-enabled Ambient Assisted Living systems. In: 2015 IEEE 2nd World Forum on Internet of Things (WF-IoT), pp. 711–714. IEEE, December 2015Google Scholar
- 32.Barnard-Wills, D., Marinos, L., Portesi, S.: Threat landscape and good practice guide for smart home and converged media. Technical report, ENISA (2014)Google Scholar
- 33.Kouzinopoulos, C.S., Spathoulas, G., Giannoutakis, K.M., Votis, K., Pandey, P., Tzovaras, D., Katsikas, S.K., Collen, A., Nijdam, N.A.: Using blockchains to strengthen the security of internet of things. In: Gelenbe, E., Campegiani, P., Czachorski, T., Katsikas, S., Komnios, I., Romano, L., Tzovaras, D. (eds.) Euro-CYBERSEC 2018. CCIS, vol. 821, pp. 90–100. Springer, Cham (2018)Google Scholar
- 34.Gelenbe, E., Kadioglu, Y.M. : Energy life-time of wireless nodes with network attacks and mitigation. In: ICC 2018 Workshops. IEEE (2018)Google Scholar
- 35.Brun, O., Yin, Y., Gelenbe, E., Murat Kadioglu, Y., Augusto-Gonzalez, J., Ramos, M.: Deep learning with dense random neural networks for detecting attacks against IoT-connected home environments. In: Gelenbe, E., Campegiani, P., Czachorski, T., Katsikas, S., Komnios, I., Romano, L., Tzovaras, D. (eds.) Euro-CYBERSEC 2018. CCIS, vol. 821, pp. 79–89. Springer, Cham (2018)Google Scholar
<SimplePara><Emphasis Type="Bold">Open Access</Emphasis> This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.</SimplePara> <SimplePara>The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.</SimplePara>