Steganography and steganalysis for digital image enhanced Forensic analysis and recommendations

Figures & data

Table 1. Signature sub-techniques overview.

Name	Advantages	Disadvantages
Most SSS	-	Knowledge of steganography tool used. Reliable when traces are left
Fridrich’s USS	Reliable	Works only with JPEG images only. Sensitive towards image re-scaling

Table 2. Statistical sub-techniques overview.

Name	Advantages	Disadvantages
Chi-square analysis	Detects reliably sequentially embedded messages	Lower accuracy for randomized embedded messages
RS	More reliable than Chi-square	If the embedded message is less than 0.005 bits per pixel, it is undetectable
RQP	Reliable	Works only with 24-bit colourful images

Table 3. AI-based sub-techniques overview.

Name	Advantages	Disadvantages
Zhang et al. [52]	High detection accuracy	Works on JPEG images only
Zong et al. [53]	95% detection accuracy Good detection against double-compressed images	Works on JPEG images only
Desai et al. [54]	97% detection accuracy	-

Table 4. Image steganography tools.

Name of the tool	Suitable for JPEG images?	Free/paid version
Hide and Seek	×	Free
S-Tools (V. 4)	×	Paid
Xiao Steganography	×	Free
F5	✓	Free
Steganos Privacy Suite	✓	Paid
Invisible Secrets	✓	Paid
Binwalk	×	Free
SteganographX Plus	×	Free
Outguess	✓	Free
Steghide	✓	Free
Crypture	×	Free

Table 5. Image steganalysis tools.

Name of the tool	Suitable for JPEG images?	Free/paid version
Aletheia	✓	Free
StegExpose	✓	Free
StegoHunt	✓	Paid
Win Hex	✓	Paid
AxiomCyber	✓	Paid

Figure 1. Steganalysis technique selection.

A diagram of selecting steganalysis techniques based on steganography method identified.

Figure 2. Sample JPEG image dataset.

A sample of the JPEG image dataset of 512× 512 image size.

Table 6. Secret messages.

Message Text (Acronym)	Size
Secret (m1)	6B
This message is secret and not supposed to be found (m2)	53B
The idea of the current research is to find out how certain steganalysis tools perform against common steganography algorithms. (m3)	125B
Now is the winter of our discontent Made glorious summer by this sun of York … (m4)	529B
Now is the winter of our discontent Made glorious summer by this sun of York; And all the clouds that lour’d upon our house In the deep bosom of the ocean buried. Now are our brows bound with victorious wreaths … (m5)	1040B

Table 7. Cover and stego images.

Tool	Cover Image Name	Cover Image	Stego Image (m1)	Stego Image (m3)	Stego Image (m5)
F5	BC
OutGuess	BG
Steghide	BC

Table 8. StegExpose detection rate.

StegExpose detection accuracy
Secret Message	OutGuess		F5		StegHide
	GS	RGB	GS	RGB	GS	RGB
m1	50%	50%	50%	50%	50%	50%
m2	50%	50%	50%	50%	50%	50%
m3	50%	50%	50%	50%	50%	50%
m4	50%	50%	50%	50%	50%	50%
m5	50%	50%	50%	50%	50%	50%

Table 9. Aletheia detection rate.

StegExpose detection accuracy
Secret Message	OutGuess		F5		StegHide
	GS	RGB	GS	RGB	GS	RGB
m1	26.6%	33.3%	46.6%	40.0%	33.35	33.3.%
m2	26.6%	33.3%	46.6%	40.0%	33.3%	33.3%
m3	26.6%	33.3%	46.6%	40.0%	33.3%	33.3%
m4	40.0%	33.3%	46.6%	46.6%	40.0%	33.3%
m5	40.0%	46.6%	60.0%	46.6%	33.3%	33.3%

Table 10. Sample stego images values.

Image name	Average Stego ImageSize in Bytes	Type	Colour	Stego Image
BC	19859.0	jpg	RGB
BG	19341.6	jpg	GS
CC	36188.0	jpg	RGB
CG	32496.4	jpg	GS
SC	45863.0	jpg	RGB
SG	42713.4	jpg	GS

Table 11. F5 stego images feature analysis results.

Image Name	Colour	Mean Size (Bytes)	SD Sizes (Bytes)	MSE-Mean	MSE-SD	Mean PSNR (dB)	SD PSNR (dB)	RMSE
BG	GS	19341.60	377.54	0.69	0.40	50.34	2.24	0.83
BC	RGB	19859.00	412.04	1.35	0.43	47.01	1.25	1.16
CG	GS	32496.40	203.42	0.80	0.34	49.40	1.59	0.89
CC	RGB	36188.00	182.61	1.93	0.44	45.37	0.91	1.38
SG	GS	42713.40	150.46	0.98	0.35	48.45	1.38	0.98
SC	RGB	45863.00	164.11	1.86	0.39	45.52	0.83	1.36

Table 12. Outguess stego images feature analysis results.

Image Name	Colour	Mean Size (Bytes)	SD Sizes (Bytes)	MSE -Mean	MSE-SD	Mean PSNR (dB)	SD PSNR (dB)	RMSE
BG	GS	18751.40	62.28	1.00	0.51	48.64	2.04	0.98
BC	RGB	19365.60	71.46	1.51	0.72	46.77	1.88	1.37
CG	GS	30787.20	33.84	1.29	0.54	47.36	1.67	1.13
CC	RGB	34424.00	40.98	2.62	0.79	44.13	1.21	1.61
SG	GS	40361.00	19.67	1.52	0.55	46.56	1.45	1.23
SC	RGB	43570.40	36.09	2.60	0.76	44.14	1.16	1.61

Table 13. Steghide images feature analysis results.

Image Name	Colour	Mean Size (Bytes)	SD Sizes (Bytes)	MSE- Mean	MSE- SD	Mean PSNR (dB)	SD PSNR (dB)	RMSE
BG	GS	27099.40	25.78	0.05	0.03	61.62	2.74	0.22
BC	RGB	19915.80	32.99	0.30	0.17	54.18	2.69	0.54
CG	GS	44737.20	21.19	0.06	0.04	61.19	2.81	0.24
CC	RGB	35839.80	12.67	0.36	0.21	53.49	2.88	0.6
SG	GS	58438.00	17.74	0.06	0.04	60.97	2.76	0.24
SC	RGB	45328.40	15.12	0.32	0.19	54.02	3.03	0.56

Mehta P, Nair A, Edappilly S, et al. A comprehensive study of AI-Based steganalysis techniques on image and text documents. In: Tiwari S, Trivedi MC, Kolhe ML, Mishra KK Singh BK, editors. Advances in data and information sciences. Lecture notes in networks and systems. Springer; 2022. p. 53–63. doi: 10.1007/978-981-16-5689-7_5.

 Google Scholar

Zhang Z, Hu D, Yang Y, et al. 2013. A universal digital image steganalysis method based on sparse representation. In: Ninth International Conference on Computational Intelligence and Security, p. 437–441. 10.1109/CIS.2013.99

 Google Scholar

Zong H, Liu F-L, Luo X-Y. Blind image steganalysis based on wavelet coefficient correlation. Dig Inv. 2012;9(1):58–68. doi: 10.1016/j.diin.2012.02.003

 Web of Science ®Google Scholar