iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Hidden data usually dies in a chat app because the app changes the pixel values that the hiding method depends on. The clearest recent measurement is a 2026 Telegram experiment: least-significant-bit (LSB) messages extracted perfectly when the image was sent through Telegram’s send-document route, but extraction failed when the same images went through the send-image route. Robust watermarks are built to survive this kind of processing, yet their results depend on the algorithm, the attack, and the route. No single finding covers every app.
Why LSB embedding breaks under recompression
LSB steganography overwrites the lowest-order bit of pixel values with bits of a hidden message. The change is too small to see, and extraction is exact: the receiver reads back the same bits that were written. That exactness is also the weakness. Any step that recalculates pixel values, such as lossy JPEG recompression or resizing, can shift those low-order bits, and a single flipped bit corrupts the message.
JPEG is lossy by design. It converts image blocks into frequency components and quantizes them, discarding detail judged less visible. The decoded image is rebuilt from those rounded values, so the original low-order bits are not guaranteed to return. This is the general mechanism behind most LSB failures in chat apps. The studies below measure its effect in specific apps and conditions; they do not measure every possible transformation.
Steganography and watermarking have different goals
The 2024 article by Pengfei Wang and coauthors, Covert Communication through Robust Fragment Hiding in a Large Number of Images, separates the two fields this way:
#1 Best Overall
- Used Book in Good Condition
“Steganography schemes mainly focus on capacity, invisibility, and security, and their protected object is confidential information; watermarking schemes mainly focus on robustness and invisibility, and their protected object is the carrier.”
In practice, LSB is a steganography technique tuned for capacity and invisibility, while a watermark is tuned to keep a signal detectable after the carrier is altered. Real designs borrow from both, which is why a comparison has to state which goal a method was built for.
What the Telegram LSB measurements show
Fitriyani and Fachri, in the Jurnal Teknologi Informasi dan Multimedia (published 25 May 2026), tested LSB messages in 15 images in PNG, BMP, and JPG formats sent through Telegram. Their results split by route:
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minute| Telegram route | LSB extraction result | Authors’ explanation and conditions |
|---|---|---|
| Send as document | 100% extraction success; bit-error rate 0 | Reported for the 15-image experiment only |
| Send as image | Extraction failed | Attributed by the authors to compression |
The contrast shows that the route matters as much as the file. A document send preserved the carrier well enough for exact extraction in this experiment; a photo send did not. The authors’ journal page also states that the experiment’s downloadable data were not yet available when the article was posted, so the result cannot yet be independently reproduced from that source. The finding is specific to Telegram. It should not be applied to WhatsApp or other messengers without a separate test.
Rank #3
How a robust watermark is measured
The 2024 Wang et al. method spreads message fragments across a large number of images, adds redundancy and coding, and uses DFT and DCT stages so that the hidden signal sits in transform-domain coefficients rather than in individual low-order bits. Its authors then test named transformations rather than claiming general robustness.
| Test condition (Wang et al., 2024) | Reported result |
|---|---|
| Rotation, scaling, cropping | Full recovery under the tested settings |
| JPEG quality factor 80 plus cropping (combined attack) | Full recovery under the tested settings |
| Up to 30% of received images lost | 100% recovery |
| 80% of received images lost | 61.1% recovery |
Image quality
The authors report an average PSNR of 41 dB for their method. This is their own figure for their own dataset and protocol. Do not place it beside PSNR values from other studies as if they were measured the same way.
Capacity and the trade-off
Distributing fragments across many images is what makes recovery possible when some images are lost, but it limits how much each image carries. The authors state that the capacity per image is limited.
Weak points
The same paper names a clear weakness in its limitations section: “Our watermarking scheme is not very robust to contrast and luminance changes.” A chat app that alters brightness or contrast along with compression would therefore be a harder case for this method than the tested attacks suggest.
Measurements that do not line up
Several other studies report numbers that look comparable at a glance but measure different things. The table sets out what each one tested.
| Source | Method | Channel or condition | Key reported figure | Limit on interpretation |
|---|---|---|---|---|
| Bunzel, Chen, and Steinebach, ARES 2022 (Fraunhofer repository record) | F5 as a proof of concept | Telegram API limits | Optimum at 2560 × 2560 pixels, JPEG quality 82, average payload of 81 KB per image | The reported optimum for that test setup, not a general Telegram specification; it does not compare LSB with a watermark |
| Wang et al., 2024 | Robust fragment hiding | Own rotation, scaling, cropping, and JPEG attacks | Full recovery under tested attacks; average PSNR 41 dB | Results apply to this algorithm and its fragment-redundancy design |
| IEEE conference paper, 2026 (abstract on IEEE Xplore) | Hybrid DCT and Reed-Solomon approach | Telegram photo mode, 50 natural HD images | Recovery reported when resolution was preserved | Sensitive to resizing; only the abstract was available for review |
| Fitriyani and Fachri, 2026 | LSB | Telegram send-document and send-image, 15 images | Perfect extraction by document; failure by image | Single app, small sample, data not yet downloadable |
Does WhatsApp or another app behave the same way?
The sources reviewed here do not establish how WhatsApp or other chat apps process photos. The Telegram LSB result applies to Telegram’s two sending routes only. No official platform specification of current image-processing behaviour was found, so each app and version has to be tested on its own. Chat apps can also change behaviour over time, which makes a measurement from one date less reliable for another.
How to compare hiding methods fairly
- Payload capacity: bytes per image, and whether the method uses one carrier or distributes fragments across many.
- Recovery after processing: extraction rate or bit-error rate after named operations, such as JPEG recompression, resizing, cropping, or rotation. One successful test is not universal robustness.
- Image quality: the metric used (for example, PSNR), its test set, and the conditions under which it was computed.
- Channel and route: the app, the sending mode (send-document, send-image, or photo mode), and whether the resolution changed.
- Secrecy and detectability: recovery is not secrecy. A method that survives transmission may still be detectable by steganalysis.
How to test your own app route
- Create a short test message and embed it with your LSB tool into a lossless PNG, recording the original message and a checksum.
- Send the image through the exact app route you intend to use, such as Telegram as a document, and save the file the recipient receives.
- Extract the message from the received file and compare it with the original, counting bit errors.
- Repeat with the other route (for example, as a photo) and with the same app version on another day, noting any change in resolution or file size.
What these measurements settle, and what they do not
The measurements support a clear mechanism: LSB messages depend on exact pixel values, and recompression can break that exactness. They also show that the route an app uses can decide whether a message survives. Robust watermarks can keep a signal through named attacks, but they trade away capacity and can fail under contrast or brightness changes. What the current evidence does not provide is a universal ranking of LSB and watermarking across chat apps, and it does not show that any single chat app treats all images the same way.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
