CC BY
12
УДК 81'33
А. В. Джунковский
специалист по УМР, аспирант кафедры прикладной и экспериментальной лингвистики Института прикладной и математической лингвистики факультета английского языка Московского государственного лингвистического университета; е-mail: [email protected]
перспективы использования современных технологий виртуальной реальности в качестве стеганографической среды
Статья посвящена анализу перспектив использования современных технологий виртуальной реальности в контексте разработки стеганографических методов лингвистической защиты информации. В последние десятилетия технологии виртуальной реальности значительно эволюционировали. Популяризация данной технологии с последующим выходом на массовый рынок привела к увеличению технической сложности программно-аппаратных элементов виртуальной реальности. В данном контексте становится возможным рассмотрение перспектив сокрытия смысловой информации в пространствах виртуальной реальности в рамках стеганографического подхода. В работе анализируются особенности современных технологий виртуальной реальности, а также сложности, сопряженные с их использованием в рамках стеганографии и стеганоанализа.
Ключевые слова: стеганография; стеганоанализ; технологии виртуальной реальности; Oculus Rift; «симуляторная болезнь».
A.V. Dzhunkovskiy
Education and Methodology Specialist, Postgraduate Student of the Department of Applied and Experimental Linguistics of the Institute of Applied and Mathematical Linguistics, Moscow State Linguistic University; e-mail: [email protected]
PROSPECTS OF USING VR TECHNOLOGIES AS A STEGANOGRAPHY MEDIUM
The present paper is devoted to the analysis of prospects of using modern VR technologies in the context of developing steganographic means of linguistic information protection. In the last decades, VR technologies have undergone rapid evolution. The popularization of these technologies in conjunction with rapidly growing sales have prompted steady growth of both software and hardware complexity of VR technologies. This makes it possible to conduct analysis of prospects of using virtual reality as a medium for the use of steganographic means of linguistic information protection. In this paper we consider the peculiarities of modern VR technologies and the difficulties associated with their use both in steganography and steganalysis.
Key words: steganography; steganalysis; VR technologies; Oculus Rift; VR-sickness.
I. INTRODUCTION
In the modern world, steganography is rapidly gaining importance alongside cryptography as a major branch of information protection in the digital age. This is correlated with the rise of demand on not only protecting sensitive information, but doing so in such manner that prevents detection of the very fact that any hidden information is being transferred [Encyclopedia 2014, p. 117-123].
One of the main advantages of using steganography as a means of information protection is that it is usually better protected against automatic and semi-automatic analysis. As such, a highly-trained human expert is required to counteract steganographic methods [Алферов и др. 2012]. In the best case scenarios, successful use of steganography would only allow said steganalysis expert to conclude that there is a probability of the text containing containers with hidden information.
Compared to cryptography, where in most cases the fact of encryption is evident, steganography provides solutions for the challenges of the modern world characterized by both mass surveillance and readily abusable cybersecurity vulnerabilities. Such a combination further raises the importance of developing scientific approaches to steganography.
Although steganography as a practical activity has existed since ancient times, only in the recent decades has the scientific method been applied to it for the first time. Since then, a somewhat sporadic, yet increasingly tentative interest towards the field has emerged in different parts of the world [Bennet 2013; Katzenbeisser 2000; Wayner 2005; Xiang, Sun, Luo 2014; Потапова 2010].
We have previously conducted work on synthesizing a classification of methods of visual steganalysis as well as of steganography methods aimed at preventing and counteracting visual attacks. The results of our research have allowed us to formulate a tri-stage steganalytical method, encompassing the meta-analysis stage in which text design and formatting is under analysis, the linguistic analysis proper stage, which deals with the analysis of linguistic features of the text, and the context and semantic analysis stage [Dzhunkovskiy 2018].
Nowadays, another phenomenon is quickly becoming a major part of IT, education and the entertainment industry: virtual reality. In this paper we analyze the prospects of using virtual reality as a steganographic medium.
II. RESEARCH
a. Background
Modern virtual reality technologies use head-mounted displays (HMD) often paired with motion controls with haptic feedback. The stereoscopic video feed for each eye is slightly angle-offset in the HMD, which creates the illusion of both immersion and of 3D perception.
There are 3 major classes of HMD technology available to consumers. The first class uses smartphones as displays and processing units. The smartphone is inserted into a special mask containing lenses that helps emulate 3D vision, whereas the gyroscopes inside the smartphones allow the user to alter the view by moving their head. This class of HMD VR technology is the most accessible and, at the same time, least immersive. Virtual reality demands very high processing power, ideally allowing for at least 45 frames per second refresh rates for each eye (90 frames per second total) in order to provide a stable experience. Even the highest tier smartphones are not possessed of enough computational power or battery life to power anything but the most rudimentary VR applications. This class includes such products as Samsung Gear VR, Google Daydream View, Google Cardboard, Merge VR Goggles and a number of others. While this class of devices requires a high-end smartphone to function, the prices of HMDs themselves can be as low as 15 USD for the Google Cardboard (price relevant for April, 2019), making them financially accessible.
The second class of HMDs is standalone VR headsets. This class is characterized by only requiring the HMD and the provided motion controls to function. The processing of the image is done internally within the HMD and the only requirement is a power source. This class is relatively novel and includes such products as Oculus Go, HTC Vive Focus and Lenovo Mirage Solo, but a number of new products, such as Oculus Quest, is being developed and is scheduled for released in 2019.
Finally, the most performance-heavy and demanding class, both hardware-wise, setup-wise and financially, is the class of tethered HMDs. These devices require to be connected to a high-end PC and additionally require the setup of an array of sensors around the desired VR space for 360 degrees motion tracking. The aforementioned sensors track both the HMD itself and the provided wireless motion controllers used in each hand. These devices also either support or include headphones and microphones
for audio feed, further improving immersion. This class of devices includes Oculus Rift, HTC Vive (and the HTC Vive Pro variation) and various Windows Mixed Reality (WMR) devices, the most prominent of which are Samsung HMD Odyssey and Lenovo Explorer. While these devices use different software platforms, every device can use any application originally developed for any other HMD of this class.
The emulated virtual scape depends on the application being used, with more and more of them supporting multi-user interaction, such as the VRChat and Rec Room applications allowing users to interact via voice over IP (VoIP) while simultaneously engaging in various forms of entertainment (e.g. playing a match of virtual tennis or visiting a gallery). These applications also support full creative freedom for the users, allowing them to model their own worlds and design their own activities within the overarching meta-world. The aforementioned applications are support by all devices in the "tethered" category.
b. Methods and Research Equipment
In order to conceptualize and analyze the prospects of using VR technologies as a steganographic medium, we have extensively tested HMDs first-hand.
The chosen HMD has been Oculus Rift CV1 (Consumer Version 1). The choice of this device was dictated by three main factors:
I. this device is in the "tethered" class, allowing access to the highest number of software applications and the highest quality of the VR simulation;
II. is the most lightweight of the devices of this class (450 grams), allowing for easier testing and less fatigue;
III. includes integrated headphones in the HMD, minimizing deviation between experiences of different users and aiding research result consistency.
As mentioned, the device requires a PC to be connected to as well as an array of motion sensors.
Oculus Rift CV1 can function with one, two or three Oculus Sensors, providing different experiences. A one-sensor setup is a legacy configuration that is capable of tracking the motions of the HMD itself, allowing the user to control the view in virtual reality with head movements.
The two-sensor setup allows for 180-degree tracking of both the HMD and two manual motions controllers with haptic feedback (Oculus Touch).
This setup is usable, yet creates a large motion control perception dead zone behind the user, potentially breaking immersion and creating application control issues.
Finally, the three-sensor setup with two sensors in front of the user at head height and one behind the user pointing down allows for the so-called "Roomscale VR", allowing full 360-degree motion control tracking and the highest supported tracking quality. This has been the chosen sensor setup for this research (Fig. 1).
Fig. 1. Oculus Rift 360-degree 3-Sensor Research Setup
The image has been extracted from the provided official Oculus Rift CV1 configuration tool. The green rectangle area represents the recommended usage area (2x1,5 meters), the light-blue areas indicate available physical space and overlocking sensor fields of vision indicate the possible coverage and overlap. For the correct functioning of the 360-degree setup, each Oculus Touch motion controller and the HMD itself must be in view of two of the three sensors at all times.
The developer provides a list of both minimal and recommended PC specifications (Table 1).
Table 1
Oculus Rift CV 1 Specifications
Minimum Recommended used research Setup
Graphics card NVIDIA GTX 1050Ti / AMD Radeon RX 470 NViDiA GTX 1060 / AMD Radeon RX 480 NViDiA GTX 1080 (Gigabyte Windforce, 8 GB GDDR5X VRAM)
Alternative graphics card NVIDIA GTX 960 / AMD Radeon R9 290 NViDiA GTX 970 / AMD Radeon R9 290
cpu Intel i3-6100 / AMD Ryzen 3 1200, FX4350 intel i5-4590 / AMD Ryzen 5 1500X intel Core i7-4770K @ 3800 MHz
Memory 8GB+RAM 8GB+RAM 16GB (DDR3, 2x Kingston 99U5403-519.A00LF)
Video output Compatible HDMI 1.3 video output Compatible HDMi 1.3 video output HDMi-2.0b video output
uSB Ports 1x USB 3.0 port + 2x USB 2.0 ports 3x USB 3.0 ports + 1x USB 2.0 port 4x USB 3.0 ports
os Windows 10 Windows 10 Windows 10 Ultimate x64
As shown in table 1, the research setup we used has exceeded the recommended PC specifications in every way. This was done intentionally in order to eliminate any chance of performance issues when testing any application in VR.
It must be noted that during the sensor setup process, the front sensors must be connected via USB 3.0 ports (connected to motherboard ports in our setup), whereas the third sensor is connected via USB 2.0 (connected via front USB 2.0 sockets with the use of a 4.8-meter USB 3.0 repeater extension cable manufactured by Monoprice). The use of a repeater (also called "active") extension cable was important as active tracking devices require highest possible fidelity to maintain constant tracking. This setup minimizes the amount and the size of tracking dead spots. While a 4-sensor setup is possible, we have encountered no issues using a 3-sensor setup and have elected to conduct testing using the described setup.
While completing the setup, we have encountered an issue: it appears that the sensors, which are simple monochrome USB cameras in nature, can malfunction when the setup area is either too bright or has too many other devices creating electromagnetic interference. This interference does not seem to affect the functioning of the devices post-setup, but must be eliminated to calibrate the VR set for first-time use.
c. Results
While conducting the research, we have used Oculus Rift CV1 every day for three months for the duration of 20-40 minutes per day. When considering our experience in the context of using VR as a steganographic medium, we must address two major aspects: effects of VR technology on the human body and psyche and the perceived potency of the technology for the purposes of steganography.
The first issues to address is the effects of VR technology on the human body and psyche. The Oculus Rift Health and Safety Guide warns that the flashing lights and patterns may trigger seizures in some users. The device is not suitable for users under the age of 13 due to it being impossible to fit the helmet around small-sized heads. Furthermore, it is recommended to take 10-15-minute breaks every half an hour of use, even if no discomfort is felt. The Guide additionally mentions the possibility of a form of motion sickness to occur, which we will return to.
The first experience of using VR technology is rather pleasant. The fidelity of the graphical components and the depth created by stereoscopic vision is such that it creates suspension of disbelief. At the same time, it is important to note that the claims that it is possible to confuse VR environments and real world are unfounded. While the immersion is great, it is not great enough. The graphical components, while impressive, are unmistakably computer graphics and are somewhat pixelated even in the best simulations. Furthermore, the contemporary VR sets cannot emulate smell or taste.
At the same time, it is important to note how well the motion controls function. Using our three-sensor setup, we have never lost tracking of our handheld Oculus Touch controllers. Another impressive feature of these controllers is that they are capable of tracking and transferring gestures from the real world into the VR simulation. It is quite intuitive to point or produce a "thumbs-up" gesture and see it replicated perfectly in VR. The main form of using finger tracking technology within Oculus Touch,
however, is the "grab" mechanics, allowing users to pick up and use objects in VR with real-life grabbing motions.
With all these elements in mind, we feel compelled to describe the VR acclimatization process. While the first-time use (around 20 minutes) was pleasant and somewhat energizing, an attempt to reenter VR for the second time during the same day for another 20-minute session after one hour of rest resulted in, in order, a slight sensation of vertigo, a slight headache, finally followed by a debilitating headache and a feeling of motion sickness.
It appears that these negative effects have to do with two factors, causing each of the problems respectively. The headaches could have been potentially caused by a calibration error: while setting up the HMD, one must adjust the interpupillary distance (IPD) setting of the lenses. Medical research indicates that wrong IPD in people wearing glasses has been known to cause migraines and tension headaches [Drummond 1987]. We assume that a related phenomenon has occurred here. An error at this stage of the calibration has emulated the effect of wearing glasses with mismatched IPD. After recalibrating the IPD, the post-use headaches have ceased.
The second phenomenon is likely caused by the mismatch of the visual stimuli received by the eyes while wearing the HMD and the vestibular system input. In other words, the body does not feel the movement observed by the VR system user, creating a sensory conflict. This phenomenon, previously observed in pilots using flight simulators and named "simulator sickness", is manifesting in VR technologies and has been called "VR-sickness" [Kemeny et al. 2017].
Having experienced the full extent of VR-sickness repeatedly, we can report that gradual moderate use of VR technologies allows the organism to acclimate to VR environments. One of the subjectively most effective remedies for this negative effect turned out to be ground ginger and ginger tea. We feel compelled to note that we have never experienced any other form of motion sickness prior to attempting to use VR. After two months of use, VR-sickness stopped occurring altogether.
The most disturbing cluster of side effects produced by the initial uses of VR include post-use perception alternations, effects on sleep, and in-VR graphical glitches we encountered. The post-use perception alterations lasted for mere days after first use and have not resurged at any point thereafter. They included two different phenomena.
The first phenomenon occurred in dim environments: when looking at one's hands, they appeared as their VR-environment simulated models. While this perception peculiarity occurred, we felt absolute certainty that we are, indeed, in the real world and not in the VR simulation as only the hands appeared slightly unreal.
The second phenomenon is likely connected with the mismatch between VR-spaces and the real space the user is in while using VR technology. For the period of at least three days after first using VR technology, while outside the VR, an anxiety of walking into invisible objects persisted. This is likely connected with the fact that it is very possible and at first even probable to forget the limitation of VR simulations and attempt to exit the area in real world which is not limited in the VR simulation, resulting in collisions with real world objects.
The sleep effects we mentioned included a significantly heightened rate of spontaneous lucid dreaming. While inside these lucid dreams, their themes and fabulae did not contain any elements connected to VR technology. It appears that being exposed to VR can stimulate lucid dreaming, but this requires additional research.
The in-VR graphical glitches mostly manifested in the software not being able to process the visual input fast enough, creating spontaneous cases of mismatch between visual feed and the position of the user. These glitches could manifest in either the VR world becoming frozen, or the horizon tilting, in some cases considerably. Both of these effects produce immediate feelings of nausea, but can largely be prevented by instantly closing your eyes when encountered. Notably, even considerably surpassing the recommended PC specifications and the effects of the Oculus Rift CV1 built-in failsafe "Asynchronous timewarp" and "Asynchronous spacewarp" technologies, designed to insert missing frames to produce a continuous image when the graphical processing unit is throttling, did not prevent us from experiencing these phenomena.
Additionally, we noted that there was a subjective improvement of the clarity of vision. However, due to the fact that we have already had 1.00 Decimal visual acuity both prior and after using the VR technology, it is difficult to determine if there has been an objective improvement in acuity.
Even after extensive use ofVR technology, it was physically impossible for us to spend more than 40 minutes inside the VR simulation without the onset of fatigue.
d. Discussion
The aforementioned usage experience allows us to produce a number of statements that can help determine the efficacy of VR as a steganographic medium.
The first statement we have been able to formulate is that the usage of VR is highly demanding in a number of factors: it requires a high level of technical prowess to install, use and troubleshoot, is restrictive financially, requires extensive acclimatization and training to use effectively, and excludes usage by children and people with a number of medical conditions.
This works in favor of VR environments as steganographic mediums. Not only is it difficult to start using VR, it is physically impossible to conduct prolonged analysis of potential steganographic containers in VR environments due to limitations of the human body and the psychophysical toll of the technology.
Furthermore, we postulate that, on a software level, VR environments can be more secure than non-VR counterparts. This is connected with that fact that if a VR-environment is located server-side and not user-side, it becomes highly problematic to decompile the application files to manually search them for steganographic elements using traditional non-VR graphical interfaces, forcing the steganalyst to enter VR, where conducting the type of work that is highly demanding intellectually to being with is additionally made much more difficult by the strain on the mind and the body.
Finally, while using VR, we have encountered a promising VR environment interaction peculiarity. This peculiarity manifests in applications where first-person view combined with head vision controls are possible. As noted before, VR scapes and physical VR use spaces correlate, but are not equivalent. At the same time, VR technology has limited ability to imitate touch though haptic feedback and physical interaction through "grab" controls, but is completely unable to imitate consistent physical collision.
Let us consider a scenario. In VR, we are standing in a 1x1 meter room. In reality, we are in a 2x2 meter room with the 1x1 room in the center of our physical space. A question arises: what happens when we walk into a wall in VR? There are two solutions for this problem: either the user "moves" the room by walking into the wall until he hits the physical wall in real world, or he moves through the wall.
А. В. Джунковский
The presented thought experiment demonstrates one of the most potent ways in which VR can accommodate steganography: containers with linguistic steganographic messages can be hidden in locations that, in real life, are highly problematic to use. In any simulated virtual environment, be it VR or non-VR, an object is created by applying textures (graphical elements) to models (geometric elements). Messages can be hidden inside walls, under the floor, within the sides of a table. Any object can potentially become a steganographic container when using this approach.
III. CONCLUSIONS
The conducted analysis shows that VR technology is a highly potent medium for protecting information through the apparatus of steganography. This is connected with both severe psychophysical limitations of using the technology, debilitating penalties for misuse and failure to install the equipment correctly, and, perhaps most importantly, the fact that VR environments allow for a way of hiding steganographic containers that is nigh-impossible to replicate in the real world.
This approach presupposes using user-created server-side VR environments and inserting steganographic containers outside of bounds of normal human object interaction. These locations include the insides of digital object models, in the void between textures.
Access to such containers can only be gained by physically moving the HMD through the object in VR and looking inside locations not normally accessible in real world interaction.
The sum of these factors demonstrates the worth of further research of VR as a steganographic medium, however we recommend utmost caution when doing so as the side effects of modern VR technologies are severely unpleasant.
список литературы
Алферов А. П. [и др.]. Основы криптографии / А. П. Алферов, А. Ю. Зубов,
А. С. Кузьмин, А. В. Черемушкин. М. : Гелиос АРВ, 2012. 480 с. Потапова Р. К. Речь: коммуникация, информация, кибернетика. М. : УРСС, 2010. 600 с.
BennetK. Linguistic Steganography: Survey, Analysis, and Robustness Concerns
for Hiding Information in Text // CERIAS Tech Report. 2013. № 4. P. 1-30. Drummond P. D. Pupil Diameter in Migraine and Tension Headache // Journal of Neurology, Neurosurgery and Psychiatry. 1987. № 50 (2). P. 228-230.
Dzhunkovskiy A. V. Стеганография: метод трехступенчатого анализа для письменных текстов русского языка // Весник Московского государственного лингвистического университета. Гуманитарные науки. 2018. Вып. 6 (797). 2018. С. 117-123.
Encyclopedia of Cryptography and Security. Second Edition // van H. C. A. Tilborg, S. Jajodia (Ed.). Berlin : Springer Science & Business Media, 2014. 1416 p.
Katzenbeisser S. Principles of Steganography. Boston : Artech House, 2000. 186 p.
Kemeny A. [at al. ]. New VR Navigation Techniques to Reduce Cybersickness // Electronic Imaging, The Engineering Reality of Virtual Reality / A. Kemeny, P. George, F. Merienne, F. Colombet. 2017. P. 48-53.
Wayner P. Strong Theoretical Steganography. Berlin : Cryptologia, 2005. 410 p.
Xiang L., Sun X., Luo G. Linguistic Steganalysis Using the Features Derived from Synonym Frequency // Multimedia Tools and Applications. 2014. № 3. P. 1893-1911.
REFERENCES
AlferovA. P. [i dr.]. Osnovy kriptografii / A. P. Alferov, A. Ju. Zubov, A. S. Kuz'min, A. V. Cheremushkin. M. : Gelios ARV, 2012. 480 s.
Potapova R. K. Rech': kommunikacija, informacija, kibernetika. M. : URSS, 2010. 600 s.
Bennet K. Linguistic Steganography: Survey, Analysis, and Robustness Concerns for Hiding Information in Text // CERIAS Tech Report. 2013. № 4. P. 1-30.
Drummond P. D. Pupil Diameter in Migraine and Tension Headache // Journal of Neurology, Neurosurgery and Psychiatry. 1987. № 50 (2). P. 228-230.
Dzhunkovskiy A. V. Steganografija: metod trehstupenchatogo analiza dlja pis'-mennyh tekstov russkogo jazyka // Vesnik Moskovskogo gosudarstvennogo lingvisticheskogo universiteta. Gumanitarnye nauki. 2018. Vyp. 6 (797). 2018. S. 117-123.
Encyclopedia of Cryptography and Security. Second Edition // van H. C. A. Tilborg, S. Jajodia (Ed.). Berlin : Springer Science & Business Media, 2014. 1416 p.
Katzenbeisser S. Principles of Steganography. Boston : Artech House, 2000. 186 p.
Kemeny A. [at al. ]. New VR Navigation Techniques to Reduce Cybersickness // Electronic Imaging, The Engineering Reality of Virtual Reality / A. Kemeny, P. George, F. Merienne, F. Colombet. 2017. P. 48-53.
Wayner P. Strong Theoretical Steganography. Berlin : Cryptologia, 2005. 410 p.
Xiang L., Sun X., Luo G. Linguistic Steganalysis Using the Features Derived from Synonym Frequency // Multimedia Tools and Applications. 2014. № 3. P. 1893-1911.
