Where does confusion collide?

In cryptography, confusion is one of the two design goals Shannon set out for a secure cipher: each bit of the ciphertext should depend on the secret key in so tangled a way that an attacker who sees the output cannot work backward to recover the key. Shannon's 1949 Communication theory of secrecy systems paired it with diffusion, where confusion hides the relationship between key and ciphertext and diffusion spreads the influence of each plaintext bit across the output. The referent is a property of the key-ciphertext mapping, realised in practice by non-linear substitution boxes, and it has nothing to do with a classification error table or a clinical state of disorientation beyond the shared ordinary word.[1]

Unit

The statistical complexity of the dependence between key and ciphertext

Recovery means

Not applicable

Measured by

Assessed through the non-linearity of substitution components and resistance to linear and differential cryptanalysis

In clinical medicine, confusion names an acute confusional state, now usually called delirium: a sudden, fluctuating disturbance of attention, awareness, and thinking, often brought on by illness, medication, or surgery, especially in older patients. Inouye and colleagues' 1990 Clarifying confusion: The confusion assessment method gave clinicians a standardised bedside instrument to detect it from features such as acute onset, inattention, disorganised thinking, and altered consciousness. The referent is a reversible brain syndrome tracked at the bedside, unrelated to a cipher's key dependence or a statistician's error table except that all three invoke the everyday idea of being muddled.[2]

Unit

Presence and severity of acute, fluctuating disturbance of attention and cognition

Recovery means

Confusion usually clears once the underlying trigger is treated, though it can persist or recur

Measured by

Detected with structured instruments such as the Confusion Assessment Method, scoring acute onset, inattention, disorganised thinking, and altered consciousness

In statistics and machine learning, a confusion matrix (or error matrix) is a square table that cross-tabulates predicted classes against true classes, so each cell counts how often items of one actual category were assigned to each predicted category. Stehman's 1997 Selecting and interpreting measures of thematic classification accuracy works through how overall accuracy, producer's and user's accuracy, and related measures are all read off this matrix. The referent is a tabular summary of a classifier's correct and mistaken assignments, distinct from a cipher's confusion property or a patient's confusional state, sharing only the word's meaning of things being mixed up with one another.[3]

Unit

Counts of predicted-versus-actual class assignments arranged in a table

Recovery means

Not applicable

Measured by

Summarised by accuracy, precision, recall, and related rates computed from the cells of the matrix