This is a difficult problem because it depends on what you mean by noise.
If you mean the search for individual attributes whose values are incorrect, then you can plot the histogram and work out some restrictions on what constitutes the actual value. You can then enforce this rule using Filter Examples to remove them.
If you mean looking for attributes that have some kind of random jitter applied to them, it would be hard to find them. Only knowing in advance what expected form of distribution you could compare with observation and do something about it. However, the action to be taken is by no means obvious.
If you want to find examples in a set of examples that are obviously different from other examples, you might consider using various outlier functions. The easiest to start is Detect Outlier (Distances) . This finds a certain amount of outliers (10 by default) based on a distance calculation using all the attributes for the examples. It creates a new outlier attribute that is set to true or false. You can then use the Filter Examples operator to remove those that are set to true.
Hope this helps, at least in the beginning.