One hindrance to the widespread adoption of mobile-agent technology is the lack of security. Security will be the issue that has to be addressed carefully if mobile agents are to be used in the field of electronic commerce. SAFER (secure agent fabrication, evolution and roaming) is a mobile-agent framework that is specially designed for the purpose of electronic commerce (Guan & Hua, 2003; Guan, Zhu, & Maung, 2004; Zhu, Guan, Yang, & Ko, 2000). Security has been a prime concern from the first day of our research (Guan & Yang, 2002; Yang & Guan, 2000). By building strong and efficient security mechanisms, SAFER aims to provide a trustworthy framework for mobile agents to assist users in conducting mobile or electronic-commerce transactions. Agent integrity is one such area crucial to the success of agent technology (Wang, Guan, & Chan, 2002). Despite the various attempts in the literature, there is no satisfactory solution to the problem of data integrity so far. Some of the common weaknesses of the current schemes are vulnerabilities to revisit attacks, when an agent visits two or more collaborating malicious hosts during one roaming session, and illegal modifi- cation (deletion or insertion) of agent data. The agent monitoring protocol (AMP; Chionh, Guan, & Yang, 2001), an earlier proposal under SAFER to address agent data integrity, does address some of the weaknesses in the current literature. Unfortunately, the extensive use of PKI (public-key infrastructure) technology introduces too much overhead to the protocol. Also, AMP requires the agent to deposit its data collected to the agent owner or butler before it roams to another host. While this is a viable and secure approach, the proposed approach, Secure Agent Data Integrity Shield (SADIS), will provide an alternative by allowing the agent to carry the data by itself without depositing them (or the data hash) onto the butler. Besides addressing the common vulnerabilities of current literature (revisit attacks and data-modification attacks), SADIS also strives to achieve maximum efficiency without compromising security. It minimizes the use of PKI technology and relies on symmetric key encryption as much as possible. Moreover, the data encryption key and the communication session key are both derivable from a key seed that is unique to the agent’s roaming session in the current host. As a result, the butler can derive the communication session key and data encryption key directly. Another feature in SADIS is strong security. Most of the existing research works focus on detecting integrity compromise (Esparza, Muñoz, Soriano, & Fomé, 2006) or bypassing integrity attacks by requiring the existence of a cooperating agent that is carried out within a trusted platform (Ouardani, Pierre, & Boucheneb, 2006). However, these works neglect the need to identify the malicious host. With SADIS, the agent butler will not only be able to detect any compromise to data integrity, but will identify the malicious host effectively.