Article Preview
TopIntroduction
Covert channel is a specific branch of data hiding which aims to deliver secret messages to the potential receivers without causing the attentions of the third parties(Zander, Armitage, & Branch, 2007). The secret messages in covert channels are always embedded in multimedia files(Cheddad, Condell, Curran, & Kevitt, 2010; Vojt, #283, Holub, & Fridrich, 2013), network packets(Gianvecchio, Wang, Wijesekera, & Jajodia, 2008; Mileva & Panajotov, 2014; Shah, Molina, & Blaze, 2006), or communication behaviors. The network covert channel is most popular type, in which network packets are used as the carrier. Secret messages are transmitted by padding some bits into the packet headers(Mileva & Panajotov, 2014) or manipulating the packet timing information(Gianvecchio et al., 2008; Shah et al., 2006). However, as the most widely-studied branch of covert channels, pattern matching(Zhai, Liu, & Dai, 2013) and some statistic-based detection tools(Fahimeh Rezaei, Hempel, & Sharif, 2017) have been applied to threaten the network covert channels.
With the development of wireless communication, wireless covert channels have begun to draw researchers’ attentions, in which the secret messages are embedded by modifying wireless communication protocol fields or wireless signals. Due to the localized transmission of wireless communication, wireless covert channels are difficult to access. The warden must set observation points in each possible hotspot, capture and analyze massive wireless signals. The detection of wireless covert channels is more difficult than the packet level analysis of network covert channels.
Earlier wireless covert channels were established by padding or modifying the redundant fields of wireless communication protocols. In several wireless covert channels, the secret messages were embedded in the padding of frames, headers of the MAC, RLC, and PDCP (Grabska & Szczypiorski, 2014; Szczypiorski & Mazurczyk, 2010), the phase of STF, the frequency of CFO and Cyclic Prefix in Wi-Fi system(Classen, Schulz, & Hollick, 2015). The subcarriers reserved in OFDM-based system can also be used to transmit secret messages(Hijaz & Frost, 2010). However, these wireless covert channels are weak to the detection methods based on matching the fields of wireless communication protocols(Fatemeh Rezaei, Hempel, Peng, Qian, & Sharif, 2013).
Due to channel interference in practical wireless environment, the distortion of the wireless signal is inherent. Later, the noise-based wireless covert channels were established by converting the secret messages into the artificial noise for transmission. There are two benches of the researches on these wireless covert channels. One is the research on the theoretical covert capacity of the wireless covert channels. Assume that the warden have full knowledge of normal channel noise and the artificial noise can be transmitted directly, the theoretical covert capacity is got while meeting the undetectability that the warden can not distinguish the normal channel noise and artificial noise by hypothesis testing and relative entropy. The limits on undetectability of those wireless covert channels was studied in (Bloch, 2015), which was extended to the network with friendly nodes producing artificial noise(Soltani, Goeckel, Towsley, Bash, & Guha, 2018) and wireless relay networks(Hu et al., 2018).
The other bench is the research on the algorithm of these wireless covert channels. The research assumes that the warden have partial knowledge of normal channel noise and the artificial noise is always added to the normal wireless signal for transmission, then the wireless covert channels are proposed and benchmarked on undetectability and reliability. The undetectability of noise-based wireless covert channels was guaranteed by bringing random noise into the generation of artificial noise. In the work(Dutta, Saha, Grunwald, & Sicker, 2012), the artificial noise was generated by dirty constellation, which composes of the signal modulated from secret messages and random noise. Later, multiplex technique is also employed to add random noise into the signal to generate artificial noise(Kitano, Iwai, & Sasaoka, 2011). In our prior work (Cao et al., 2018), the artificial noise modulated from secret messages by constellation shaping modulation can maintain the same distribution as that of normal channel noise. Although the state-of-the-art schemes have achieved good performance on undetectability, the artificial noise is vulnerable to channel interferences owing to its dense distribution.