The Internet of Things (IoT) is an emerging technology, which comprises wireless smart sensors and actuators. Nowadays, IoT is implemented in different areas such as Smart Homes, Smart Cities, Smart Industries, Military, eHealth, and several real-world applications by connecting domain-specific sensors. Designing a security model for these applications is challenging for researchers since attacks (for example, zero-day) are increasing tremendously. Several security methods have been developed to ensure the CIA (Confidentiality, Integrity, and Availability) for Industrial IoT (IIoT). Though these methods have shown promising results, there are still some security issues that are open. Thus, the security and authentication of IoT based applications become quite significant. In this paper, we propose TD2SecIoT (Temporal, Data-Driven and Dynamic Network Layer Based Security Architecture for Industrial IoT), which incorporates Elliptic Curve Cryptography (ECC) and Nth-degree Truncated Polynomial Ring Units (NTRU) methods to ensure confidentiality and integrity. The proposed method has been evaluated against different attacks and performance measures (quantitative and qualitative) using the Cooja network simulator with Contiki-OS. The TD2SecIoT has shown a higher security level with reduced computational cost and time.