Abstract
This work examines the general problem of detecting the presence of hardware keyloggers (HKLs), and specifically focuses on HKLs that are self-powered and take measures, such as passively tapping the keyboard line, to avoid detection. The work is inspired by the observer effect, which maintains that the act of observation impacts the observed. First, a model for HKLs is proposed, and experimentally validated, that explains how attaching a HKL necessarily affects the electrical characteristics of the system it is attached to. The model then motivates the selection of features that can be used for detection. A comparison framework is put forth that is sensitive enough to identify the minute changes in these features caused by HKLs. Experimental work carried out on a custom keylogger designed to conceal its presence, at the expense of reliability, shows that it is possible to detect stealthy and evasive keyloggers by observing as few as five keystrokes. Optimal attack strategies are devised to evade detection by the proposed approach and countermeasures evaluated that show detection is still possible. Environmental effects on detection performance are also examined and accounted for.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
Notes
- 1.
In Sect. 5.2 we do examine the case of an evasive HKL designed to defeat our detection method by reproducing the keyboard’s signal exactly.
- 2.
In Sect. 5.3 we show that HKLs that do not affect line voltage—i.e. those with high input impedance—can still be detected because of their affect on the transient response of the system.
- 3.
Properly speaking, we use a variant of the EMD for non-normalized histograms, where we have selected the \(l_1\) norm for the ground distance metric [27].
- 4.
A slight change was made to our experimental setup to accommodate the duration of the data runs. Instead of the space bar being manually pressed, a program was written that toggled the NUMLOCK state. Since the OS state of this key and the NUMLOCK LED must be consistent, the PC would signal the keyboard that it had a scancode to send by bringing the clock line low, which would then cause the keyboard to generate a clock signal that we were able to subsequently capture.
- 5.
We note that while 125 MS/s ADCs are more expensive than the 1 MS/s variety, they can still be had for less than $15, e.g. the LTI LTC2251 [23].
References
ABC News: Former Cal State student gets year in prison for rigging campus election (2013). http://abcnews.go.com/US/cal-state-student-year-prison-rigging-campus-election/story?id=19682401
Analog Devices: AD7265 Differential/Single-Ended Input, Dual 1 MSPS, 12-Bit, 3-Channel SAR ADC (2006), datasheet
Chahrvin, S.: Keyloggers–your security nightmare? Comput. Fraud Secur. 2007(7), 10–11 (2007)
Chapweske, A.: The ps/2 mouse/keyboard protocol (2003). http://www.computer-engineering.org/ps2protocol
Danev, B.: Physical-layer Identification of Wireless Devices. Ph.D. thesis, ETH Zurich, Zurich, Switzerland (2011)
Danev, B., Luecken, H., Capkun, S., Defrawy, K.E.: Attacks on physical-layer identification. In: Proceedings of the Third ACM Conference on Wireless Network Security (WiSec 2010), pp. 89–98. ACM, New York (2010)
Danev, B., Zanetti, D., Capkun, S.: On physical-layer identification of wireless devices. ACM Comput. Surv. (CSUR) 45(1), 6 (2012)
Daniels, T.E., Mina, M., Russell, S.F.: A signal fingerprinting paradigm for physical layer security in conventional and sensor networks. In: Proceedings of the International Conference on Security and Privacy for Emerging Areas in Communnication Networks (SecureComm), pp. 219–221. IEEE Computer Society (2005)
Edman, M., Yener, B.: Active attacks against modulation-based radiometric identification. Technical report, Rensselaer Polytechnic Institute, Department of Computer Science (2009), technical Report
Erbskorn, J.W.: Detection of Intrusions at Layer ONe: The IEEE 802.3 normal link pulse as a means of host-to-network authentication A preliminary performance analysis and survey of environmental effects. Master’s thesis, Iowa State University, Ames, IA (2009)
Gerdes, R., Mina, M., Russell, S., Daniels, T.: Physical-layer identification of wired ethernet devices. IEEE Trans. Inf. Forensics Secur. 7(4), 1339–1353 (2012)
Gerdes, R.M.: Physical layer identification: methodology, security, and origin of variation. Ph.D. thesis, Iowa State University, Ames, IA (2011)
Gerdes, R.M., Daniels, T.E., Mina, M., Russell, S.F.: Device identification via analog signal fingerprinting: a matched filter approach. In: Proceedings of the Network and Distributed System Security Symposium (NDSS). The Internet Society (2006)
Greene, M., Parker, M.: Method and system for detecting a keylogger that encrypts data captured on a computer, 25 July 2006, US Patent App. 11/492,581
IEEE: Standard for transitions, pulses, and related waveforms (2011), IEEE Std 181–2011
Karim, N., Agrawal, A.: Plastic packages electrical performance: reduced bond wire diameter. In: NEPCON WEST, pp. 975–980 (1998)
KeeLog: Open source DIY hardware keylogger (2012). http://www.keelog.com/diy.html
KeeLog: Keygrabber Module (2013). http://www.keelog.com/
KeyCarbon: Keycarbon Raptor (2012). http://www.keycarbon.com/
KeyCarbon: Keycarbon PCI (2013). http://www.keycarbon.com/
Kullback, S., Leibler, R.A.: On information and sufficiency. Ann. Math. Stat. 52, 79–86 (1951)
Linear Technology: LT1793 JFET Input Op Amp (1999), datasheet
Linear Technology: LTC2251/LTC2250 ADCs (2005), datasheet
Mihailowitsch, F.: Detecting hardware keyloggers, November 2010. https://deepsec.net/docs/Slides/2010/DeepSec_2010_Detecting_Hardware_Keylogger.pdf. [DeepSec 2010 Presentation]
Nakra, B.C., Chaudhry, K.K.: Instrumentation Measurement and Analysis. McGraw-Hill Education (India) Pvt Limited (2009)
Nilsson, J.W., Riedel, S.: Electric Circuits. Prentice Hall, Upper Saddle River (2010)
Pele, O., Werman, M.: A linear time histogram metric for improved SIFT matching. In: Forsyth, D., Torr, P., Zisserman, A. (eds.) ECCV 2008, Part III. LNCS, vol. 5304, pp. 495–508. Springer, Heidelberg (2008)
Salkind, N.: Encyclopedia of Research Design. SAGE Publications, Thousand Oaks (2010)
Sapra, K., Husain, B., Brooks, R., Smith, M.: Circumventing keyloggers and screendumps. In: 2013 8th International Conference on Malicious and Unwanted Software: “The Americas” (MALWARE), pp. 103–108, October 2013
Texas Instruments: High Speed Analog Design and Application Seminar: High Speed PCB Layout Techniques (2004), presentation
Texas Instruments: LM35 Temperature Sensors (2013), datasheet
Texas Instruments: Tiva TM4C123GH6PM microcontroller (2013), datasheet
Texas Instruments: Use conditions for 5-v tolerant gpios on Tiva C series TM4C123x microcontrollers (2013), application Report
The New York Times: Credit card data breach at Barnes & Noble stores (2012). http://www.nytimes.com/2012/10/24/business/hackers-get-credit-data-at-barnes-noble.html?_r=3&
Zaitsev, O.: Skeleton keys: the purpose and applications of keyloggers. Netw. Secur. 2010(10), 12–17 (2010)
Acknowledgements
The authors would like to thank Li Yin and Heidi Harper of Utah State University for their assistance in collecting data.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Appendix: Optimal Selection of HKL Input Resistance
Appendix: Optimal Selection of HKL Input Resistance
The attacker seeks to minimize the difference between the line voltage with and without the HKL in order to evade the level-based detection approach, while simultaneously minimizing the time constant associated with the HKL to lessen the increase of the rise/fall times of the clock signal. The former goal can be realized by choosing \(R_{kl} \gg R_{pc}\) to ensure that \(R_{eq}=R_{kl}\parallel R_{pc} = R_{pc}\). This, however, is achieved at the expense of the latter goal, as the time constant \(R_{eq}C_{kl}\) can only be decreased by selecting \(R_{kl}\) such that \(R_{eq} < R_{pc}\), due to the fact that the HKL capacitance is fixed. The minimum value of \(R_{eq}\), and by extension the optimal input impedance of the HKL, necessary to evade the level-based approach while minimizing the time constant of the HKL is calculated as follows.
Allow r to represent the minimum resolvable voltage drop of the ADC employed in the detector. Evading the level-based detection approach requires \(V_l-V'_l = r\), where r may be expressed in terms of the quantities controllable and/or known by the attacker as
Defining
and rearranging terms yields
Furthermore, manipulation of (5) gives
By substituting (6) into (7) we arrive at
\(\square \)
Rights and permissions
Copyright information
© 2015 Springer International Publishing Switzerland
About this paper
Cite this paper
Gerdes, R.M., Mallick, S. (2015). Physical-Layer Detection of Hardware Keyloggers. In: Bos, H., Monrose, F., Blanc, G. (eds) Research in Attacks, Intrusions, and Defenses. RAID 2015. Lecture Notes in Computer Science(), vol 9404. Springer, Cham. https://doi.org/10.1007/978-3-319-26362-5_2
Download citation
DOI: https://doi.org/10.1007/978-3-319-26362-5_2
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-26361-8
Online ISBN: 978-3-319-26362-5
eBook Packages: Computer ScienceComputer Science (R0)