An Empirical Study of Steganography and Steganalysis of Color Images in the JPEG Domain

Abstract

This paper tackles the problem of JPEG steganography and steganalysis for color images, a problem that has rarely been studied so far and which deserves more attention. After focusing on the 4:4:4 sampling strategy, we propose to modify for each channel the embedding rate of J-UNIWARD and UERD steganographic schemes in order to arbitrary spread the payload between the luminance and the chrominance components while keeping a constant message size for the different strategies. We also compare our spreading payload strategy w.r.t. two strategies: (i) the concatenation of the cost map (CONC) or (ii) equal embedding rates (EER) among channels. We then select good candidates within the feature sets designed either for JPEG or color steganography. Our conclusions are threefold: (i) the GFR or DCTR features sets, concatenated on the three channels offer better performance than ColorSRMQ1 for JPEG Quality Factor (QF) of 75 and 95 but ColorSRMQ1 is more sensitive for QF=100, (ii) the CONC or EER strategies are suboptimal, and (iii) depending of the quality factors and the embedding schemes, the empirical security is maximized when between 33% (QF=100, UERD) and 95% (QF=75, J-UNIWARD) of the payload is allocated to the luminance channel.

Steganography by Minimizing Statistical Detectability: The cases of JPEG and Color Images

“…• "Color channels Fixed repartition" (CCFR) , referred to as "Arbitrary repartition of the payload between the 3 channels" in [21], consists in setting a fraction γ of the payload R in bits per non-zero AC coefficient (bpnzAC) to be embedded chrominance channels ; The payloads in color channels (in bpnzAC) are thus determined by setting:…”

The ALASKA Steganalysis Challenge

“…We consequently decided to modify the classical embedding scheme by spreading the payload size between the different components. The question related to the best way to spread a payload of P bits among the channels Y , C b and C r for 4:4:4 sampling ratios is investigated in [36] from a practical perspective. The authors propose to tune a parameter β ∈ [0; 1] that balances the payload allocated to the luminance and chrominance channels w.r.t.…”

“…For example, β = 1 implies that all the payload is conveyed by the two chrominance components and β = 0 implies that the payload is embedded in the luminance component only. More specifically, given P Y , P C b and P C r the payload sizes for each channel, the authors of [36] propose to use the following relations:…”

“…Note that this change tends to allocate a larger payload into chrominance channels and consequently decreases the practical security. For example, for a typical image at QF75 where N Y = 10000, N C b = 1000, N C r = 1000, if we set P = 1000 bits, β = 0.4 we obtain P Y = 882 bits and P C b = P C r = 59 bits using spreading presented in [36] vs P Y = 600 bits and P C b = P C r = 200 bits for our implementation.…”

“…Practical security assessment: The payload size was evaluated empirically based on basic and fast steganalytic tests. The experiments were conducted using the DCTR (Discrete Cosine Transform Residual) [19] features set concatenated from each color channels [36] together with the low-complexity linear classifier from [4] (LCLC). This choice was based on the need to get as many payload evaluation as possible in the least amount of time.…”