Filtersteg

Filtersteg is a class of steganographic techniques that conceal information by manipulating the outputs of digital filters applied to a host medium, such as an image, audio, video, or sensor data. The method relies on the observation that filter operations produce residuals and frequency components that can be adjusted within perceptual or statistical bounds without noticeably degrading quality. Filtersteg aims to hide a payload in these filter responses rather than in raw samples alone.

In typical filtersteg schemes, a sender selects a filter bank or a set of convolution kernels and a coding rule that maps message bits to small changes in the filtered signals. The changes are constrained to preserve perceptual or statistical similarity to the original media. After embedding, the host media is transmitted or stored as a stego object. A receiver, who shares the filter set and any necessary keys, applies the same filters and decodes the message by analyzing the resulting residuals or coefficient changes in a predefined way.

The technique can operate in various domains. In the spatial or time domain, linear or nonlinear filters modify sample groups and embed data in their responses. In the frequency or transform domain, filters correspond to bands or subbands (for example, DCT, wavelet, or FIR filter banks) and data is placed in the corresponding coefficients. Robust variants seek resilience to compression, noise, or lossy transmission, often trading off capacity for reliability.

Limitations include sensitivity to filtering or recompression by third parties, potential detectability by statistical steganalysis, and the need for proper synchronization between sender and receiver. See also steganography, watermarking, and residual-based embedding methods.