Security software vendor Barracuda Networks has unveiled a technology which it claims can improve protection against image-based spam.

Image spam generally embeds text inside images to hide the content from text rule-processing layers within spam filters.

Many firewalls now ship with an optical character recognition (OCR) engine but, in early attempts to foil the technology, senders of image spam used animated GIF files to spread content across multiple frames.

The spammers also attempted to obfuscate text inside the image by corrupting it with lines, speckles and other extraneous details.

Barracuda's third-generation image spam defence has incorporated a multi-pass OCR engine which claims to offer deeper analysis by performing image pre-processing operations to normalise the images prior to final OCR passes.

The additional processing is effective against attempts to hide text by varying colour and contrast combinations of text and backgrounds.

Barracuda claims that it can achieve 95 per cent accuracy in identifying and blocking image spam.

"While much of the attention last year was on growth in the volume of image spam, we are noticing an increase in the sophistication of image spam techniques," said Stephen Pao, vice president of product management at Barracuda.

"Since last July when we announced our image spam fingerprinting and OCR defences, we have continued to enhance our defence layers to stay ahead of the latest trends."

• Image-based spam defeating filters