Detection of Abnormal Item Based on Time Intervals for Recommender Systems

Table 1

The features of the attack models.

Attack model	𝐼^𝑆 (selected items)	𝐼^𝐹 (filler items)	𝐼^𝑇 (target items)
Random attack	∅	𝑟𝑖^𝐹 = random ratings	𝑟𝑖^𝑇=𝑟max/𝑟min
Average attack	∅	The ratings for 𝑖^𝐹 distributed the mean of each item	𝑟𝑖^𝑇=𝑟max/𝑟min
Bandwagon attack	Widely popular items,; 𝑟𝑖^𝑆=𝑟max	𝑟𝑖^𝐹 = random ratings	𝑟𝑖^𝑇=𝑟max/𝑟min
Segment attack	Similar items to target items,; 𝑟𝑖^𝑆=𝑟max	𝑟𝑖^𝐹=𝑟max/𝑟min	𝑟𝑖^𝑇=𝑟max/𝑟min